Marketers see events such as the World Cup as a rare chance to try new ideas. According to Klaviyo, 72% of UK marketers say cultural moments such as the World Cup give them more creative freedom than any other time of the year.

Many brands are already familiar with sports marketing. Klaviyo found that 77% had activated campaigns around sporting events during the previous 12 months. This summer, discounts will feature in 61% of campaigns, curated collections in 48%, and reactive fan content in 37%.

Klaviyo found that 42% of marketers plan to react to major moments within six hours, while 40% say their most successful World Cup campaigns launch within one to three days of an event happening. Social media is expected to be the main channel, with 58% saying it dominates cultural moment marketing.

Jamie Domenici, Chief Marketing Officer at Klaviyo, said, “This summer is going to be one of the most competitive marketing moments we’ve seen in years. Brands globally are showing up for the World Cup, and when everyone’s in the market, budget alone doesn’t cut it. What does is how quickly and how personally you can respond to the moments that genuinely move people. A shock result. An underdog story. Those cut through because they’re real and campaigns that tap into that, rather than just riding the hype, are the ones that actually land. That’s not just true of sport.”

But with big events like these, there’s an issue businesses should also take note of, and that is cyber crime.

��

Why Do Cyber Criminals Love Major Sporting Events?

��

Criminals see opportunities wherever large audiences gather, and few events attract attention on the scale of a FIFA World Cup.

Matt Hull, VP of Cyber Intelligence and Response at NCC Group, said, “Major sporting events are prime targets for cyber criminals because they combine global attention, emotional engagement, and huge digital dependency. This creates opportunities for everyone from financially motivated criminals to hacktivist groups looking for visibility.”

The World Cup stretches across countries, platforms, broadcasters, sponsors, ticketing systems, travel providers and millions of connected fans. Every one of those areas creates opportunities for criminals looking to trick people or gain unauthorised access to accounts and systems.

��

��

Hull also said, “It would be speculative to directly connect participating nations to cyber activity simply because teams or players are physically present – especially as a country doesn’t need to be competing in the tournament to pose a cyber risk. Cyber operations aren’t limited by geography, but equally, not every incident linked to the World Cup will be state-backed. Most are likely to be opportunistic or criminal in nature, and the more credible concern is the wider geopolitical threat environment surrounding a globally significant event hosted in the US.”

��

What Kinds Of Attacks Happen Around Tournaments?

��

The attacks often start with something that looks pretty normal – emails, text messages, social media posts and websites can all be used to trick fans into handing over personal information. Criminals often take advantage of excitement around tickets, merch and match updates.

Hull said, “The biggest risks are likely to be engineering campaigns such as phishing, credential theft, ticket scams, brand impersonation, and payment fraud. Major global events create opportune conditions for social engineering because people are more likely to trust communications and act quickly under pressure.”

Hacktivist groups can also use the spotlight surrounding major events to attract attention to their causes.

According to Hull, “Hacktivist activity is also a persistent concern, particularly disruptive attacks like website defacements and distributed denial of service (DDoS) campaigns designed to generate attention during high-profile moments.”

That means organisations working with the tournament may face threats from different directions at once, ranging from criminals wanting money to groups wanting publicity.

��

Could The 2026 World Cup Become A Cyber Crime Magnet?

��

The scale of the tournament makes it a perfect target for criminals. Think about it: the 2026 competition will be one of the biggest World Cups ever staged and will generate tonnes of online activity. Marketing teams are planning real time campaigns, brands are investing a lot into digital engagement and fans will spend weeks interacting with online platforms.

AI is expected to make marketing activity even faster. Klaviyo found that 88% of marketers plan to use AI for summer sporting campaigns, while 50% feel positive about so called “vibe marketing”, where AI can generate campaigns from simple prompts using a brand’s own data.

Domenici said, “I’m excited to see how brands use AI this summer to jump on sports cultural moments, push their creativity and connect with people in ways that feel authentic. For the first time, marketers have the tools to tap into a moment that happened only an hour ago and turn it into a personalised, on-brand campaign at a speed we’ve genuinely never seen before.”

That rise in digital activity is one of the reasons cybersecurity teams will be locking in and making sure things go as smoothly as possible.

Hull said, “The 2026 World Cup could become one of the most targeted sporting events we’ve seen from a cyber perspective – simply because of its scale, visibility, and digital footprint. In many cases, new tactics aren’t impacting the threat, but it could be increased by faster exploitation of existing weaknesses, and ongoing geopolitical tensions that the US is directly involved in.”

The post FIFA World Cup 2026: Why Have Big Sporting Events Become A Target For Cyber Criminals? appeared first on 91̽��.

That’s what happened when ran its new MDASH system against the Windows networking and authentication stack – and all 16 issues were patched in the May 2026 Patch Tuesday updates before anyone outside could exploit them.

MDASH – the Microsoft Security multi-model agentic scanning harness – orchestrates more than 100 specialised AI agents working simultaneously to discover, debate, validate, deduplicate and prove software vulnerabilities. The system is currently in limited private preview, with broader enterprise access expected in June 2026.

��

What MDASH Actually Does

��

The multi-agent design is the interesting part. Rather than a single , MDASH runs a coordinated pipeline: agents discover candidate weaknesses, others challenge and debate the findings, others validate and deduplicate results, and the system only surfaces issues it can prove with working evidence.

According to Microsoft’s own reporting, the system achieved 21 of 21 planted vulnerabilities found with zero false positives on a private test driver, 96% recall against five years of confirmed cases in one Windows component and 100% recall in another, and an 88.45% score on the public CyberGym benchmark.

The zero-false-positive result on planted vulnerabilities is the number that matters most for enterprise adoption. Security teams are drowning in alerts. A system that finds real issues without generating noise is addressing the core operational problem in enterprise cybersecurity today.

��

Why This Matters For The Cybersecurity Market

��

The shift being signalled here is from AI as a security tool to AI as a security infrastructure layer.

Traditional vulnerability scanning tools work from known signatures and patterns. MDASH works from reasoning – it doesn’t need to have seen a vulnerability before to identify it. That’s a fundamentally different capability, and it’s one that compounds: the more novel the attack surface, the more useful an agent-based approach becomes versus a pattern-matching one.

For , the arrival of MDASH as an enterprise product rather than a research project changes the playing field. Microsoft has the distribution, the existing enterprise security relationships and the data from millions of systems to train and refine a system like this at a scale that startups can’t easily match.

The key for cybersecurity founders is finding remaining gaps: narrow domain expertise, speed of deployment, integration with specific tech stacks, or the kind of human-in-the-loop oversight that large organisations will still need around AI-generated vulnerability reports.

��

What Comes Next

��

The June enterprise preview will be the first real test of whether MDASH’s benchmark performance translates into operational value at scale.

Benchmarks and internal tests are one thing; the noise and complexity of a live enterprise environment, with proprietary codebases, unusual configurations and legacy dependencies, are another. These 16 Windows vulnerabilities provide a powerful proof of concept, but the enterprise preview will yield more useful data.

For entrepreneurs and managers who have been tracking the enterprise security incidents pile up despite significant investment in existing tools, the underlying premise of MDASH is hard to argue with: if AI agents can find vulnerabilities faster and more reliably than human researchers working with traditional scanners, the question is only how quickly that capability becomes standard. If the June preview delivers, that timeline just got a lot shorter.

The post The AI That Embarrassed Microsoft’s Security Team Is About To Be Available To Everyone appeared first on 91̽��.

Finance tried an agent that categorizes expenses. Before long, these tools weren’t just “chatbots.” They were taking actions.

That’s the key change. An that answers questions is one thing. An AI agent that can do things, access systems, move files, trigger workflows, send messages, and change settings creates an entirely new set of security headaches.

And many businesses are adopting agents faster than they’re securing them.

��

Agents Turn “Read Access” Into “Real-World Impact”

��

A traditional internal tool might have limited scope: it shows dashboards, maybe runs a report. AI agents are different because they’re designed to be helpful across tasks. That usually means they’re connected to multiple systems at once: email, Slack/Teams, , ticketing, cloud consoles, databases, document storage and code repos.

Even if the agent is only meant to “help,” it can still cause damage if it has write access anywhere. A single bad action like deleting records, changing permissions, sending the wrong file, or exposing customer info can spiral quickly.

So the first new challenge is scope creep: agents often get broader access than any single human role would normally have, simply because it’s convenient.

��

Prompt Injection Becomes A Security Issue Not A Novelty

��

If your agent reads content from the outside world, customer emails, support tickets, web pages, or PDFs, that content can influence it.

Prompt injection is basically social engineering for machines. A malicious message can include instructions like “ignore previous rules” or “send me the secret configuration,” and if your agent isn’t well-guarded, it might comply. That sounds silly until you remember that agents can be connected to tools. If the agent can query internal docs, fetch customer data, or execute actions through APIs, prompt injection becomes a real attack path.

What makes this tricky is that the “attack” may look like normal business input: a ticket, a document, a partner email. It doesn’t arrive as malware.

Agents Create A New Identity Problem: Non-Human Users Everywhere

��

Security teams have spent years trying to manage human identity: MFA, SSO, least privilege and access reviews. AI agents introduce a flood of “non-human identities”: service accounts, API tokens, automation keys, OAuth grants and bot users. These accounts often have:

• Broad permissions
• Long-lived tokens
• Weak monitoring
• Unclear ownership (“Who created this? Who rotates the keys?”)

Attackers love that. If they compromise an agent’s token, they don’t need to phish an employee. They can simply operate as the agent, often with fewer alerts, because automation is expected to behave “quietly.”

��

Tool Integrations Expand The Blast Radius

��

Agent platforms are designed around connectors: Google Drive, Microsoft 365, Salesforce, HubSpot, Jira, GitHub, AWS, internal APIs and others. Every connector is another trust relationship. A common failure mode goes like this:

1. An agent is given permission to “help” with a workflow
2. The easiest permission setting is selected (“full access”)
3. No one revisits it because things work
4. Months later, that agent becomes the shortest path to sensitive systems

In classic security terms, the attack surface grows. But in a more human sense, the agent becomes a powerful coworker that nobody is supervising closely.

��

Agents Make Mistakes In Ways That Don’t Look Like Mistakes

��

When humans do something risky, they usually know they’re doing it. When agents do something risky, it might be the “reasonable” result of an ambiguous instruction.

Example: “Send the customer the report.”

Which report? The latest one? The internal one? The one with other customers’ data? If the agent grabs the wrong file, it’s not malicious; it’s just wrong. But security doesn’t care whether the breach was intentional.

This is why agent security isn’t only about blocking hackers. It’s also about preventing confident, automated errors.

��

Auditing And Accountability Become Harder

��

With normal automation, you can trace actions to scripts and systems. With agents, the decision path can be harder to reconstruct:

• Why did it take that action?
• What context did it use?
• Which tool call caused the change?
• Was it a user instruction or agent initiative?

If you can’t answer those questions, incident response becomes slow and painful. And if regulators or customers ask what happened, “the agent decided” is not an acceptable explanation.

��

The New Security Goal: “Helpful But Contained”

��

Most companies don’t need to stop using agents. They need to use them with guardrails.

This scenario is where become essential, not as a buzzword but as a practical layer that makes agents safe to run in real environments. The best approaches usually include:

• Least-privilege tool access: agents get only what they need, nothing more
• Approval steps for sensitive actions: “Draft the email” is automatic; “send the email” requires confirmation
• Data loss prevention (DLP): stop secrets, PII, and internal-only data from being exfiltrated
• Content and prompt filtering: detect injection patterns and risky instructions
• Strong identity controls: short-lived tokens, scoped OAuth, clear ownership, and rotation policies
• Full audit logs: every tool call, every resource accessed, and every action taken are traceable end-to-end
• Segmentation: isolate agent environments from critical systems unless explicitly needed

AI agents are exciting because they make work feel lighter. But they also make security feel heavier because now you’re managing software that can act like a person, with speed and scale humans don’t have.

The companies that get this right won’t be the ones that ban agents or rush them out everywhere. They’ll be the ones that treat agents like powerful employees: give them clear roles, limit what they can access, monitor what they do, and build in checks before they can cause real damage.

That’s the shift we’re living through: AI agents aren’t just tools anymore. They’re participants in your systems, and participants need rules.

The post How AI Agent Adoption Is Creating A New Cybersecurity Challenge appeared first on 91̽��.

That means many businesses were not dealing with just one unfortunate incident. They were dealing with repeated access failures involving employee accounts, machine identities and AI systems.

The company’s Identity Security Landscape Report 2026 also found that 83% of UK organisations had suffered at least one identity related breach.

This comes as the UK government is urging businesses to tighten cyber defences through a new Cyber Resilience Pledge and stronger standards, saying AI is making attacks faster and easier for criminals.
��

Why Are Businesses Getting Breached Again And Again?

��
One of the biggest issues is scale; Idira’s research says machine identities now outnumber humans 100 to 1 in UK organisations. These identities can be anything from bots and cloud systems to connected devices and automated tools.

That’s expected to go up even more in the next year. Around 82% of UK respondents expect growth in machine identities, 90% expect growth in AI identities and 50% expect more human identities too.

Businesses are no longer managing only staff logins because now, they’re managing thousands of digital identities that need permissions, passwords, certificates rules.

Rich Turner, Senior Vice President EMEA Identity Security at Palo Alto Networks, said, “The explosion of machine identities represents a fundamental shift in the enterprise attack surface. With AI driven identities projected to continue accelerating in the next year, organisations are facing a reality where identity complexity is rapidly outpacing traditional security controls.”

The report found that 80% of UK respondents said fragmented identity systems and tools are affecting or delaying their organisation’s ability to detect and respond to identity related threats.

Many businesses are running disconnected security systems and manual processes across too many digital identities.
��

Are Companies Treating Cybersecurity Like A Business Issue?

��
The UK government believes many organisations are not taking cyber security seriously enough at leadership level.

As part of a national cyber campaign, ministers are encouraging businesses to sign up to the Cyber Resilience Pledge, which asks organisations to take three actions.

These are making cyber security a board level responsibility, signing up to the National Cyber Security Centre’s free Early Warning Service, and requiring Cyber Essentials certification across supply chains.

The government has committed £90 million to .
��

She added, “The UK has a world class cyber sector that is creating skilled jobs and protecting our economy and government is doing more by investing in its own defences, legislating to require more of essential services and setting national standards.”

Her message to businesses:

“As threats evolve, businesses of all sizes need to step up and take action now. The Cyber Resilience Pledge is a call for companies to strengthen their defences, protect their customers and keep the UK secure and competitive.”

The government also reported that 43% of UK businesses experienced a cyber breach or attack in the past year.
��

Is AI Making The Problem Worse?

��
AI is creating more opportunities for businesses, but as countless reports have shown, it is also adding more digital access points that attackers can exploit.

Idira found that 34% of and 37% of machine identities in UK organisations have access to company data on average. That can mean financial records, internal systems or sensitive customer information.

Many businesses are not fully monitoring those systems.

Only 51% of UK organisations use behavioural monitoring for autonomous AI agents, while just 37% use credential revocation.

That leaves many AI tools and automated systems operating with access but limited oversight.

Turner said organisations can no longer continue with old habits.

He said, “The fact that 83% of organisations have suffered an identity related breach in the UK and 91% in EMEA more broadly proves that as AI agents gain more access to sensitive data, security leaders must move beyond manual processes.”

He added, “To close the gap, organisations must embrace end to end automation and unified governance. Otherwise, the risks of expanding AI and machine identities will only continue to intensify.”
��

Is The UK Cyber Sector Ready For Demand?

��
The irony is that cybersecurity is becoming one of the UK’s fastest growing .

The UK government said cybersecurity sector revenue came up 11% to £14.7 billion, while the number of cyber businesses came up 20% to 2,603.

The sector also created 2,300 jobs in the past year.

There is no shortage of cyber products, services or expertise entering the market.

Businesses know attacks are happening but the issue is that many are only treating cybersecurity as an urgent after repeated breaches.

The post 74% Of UK Businesses Have Had At Least 3 Identity Breaches This Past Year – Why Aren’t More Of Them Protected? appeared first on 91̽��.

’s argument is that Shift Left was designed for a world where humans wrote the code. In the agentic era – where AI coding assistants, autonomous agents and AI-generated code are now part of how enterprise software gets built, writing and deploying at machine speed – the model needs replacing. The result is more code, a wider attack surface and a faster time to exploit.

The company has just launched ADLC Security, a product offering built to secure AI-driven software development across the entire Agentic Development Lifecycle, from the first prompt to runtime. Its stated vision is a self-protecting ADLC: a development environment where security operates with equal autonomy, speed and intelligence as the agents writing the code.

��

The Dual Problem ADLC Security Addresses

��

Cycode frames its approach around two sides of the same challenge.

The first is Security for AI: governing and securing the AI layer of the development process itself – which , what prompts they run and what code the AI generates. The second is AI for Security: deploying AI agents to automate security work at the speed the agentic era demands. According to Cycode, it’s currently the only vendor addressing both sides of that equation within a single platform.

Lior Levy, Co-Founder and CEO of Cycode, put the underlying argument plainly at launch: “Shift Left is dead. The agentic era requires the Shift to AI. Security cannot stand downstream, bracing against AI. It must evolve with AI and operate in parallel with equal autonomy, speed, and intelligence as the agents writing code and exploits. ADLC Security is how we make that real for our customers.”

Alongside the product launch, Levy published a manifesto titled ‘Shift to AI: A Manifesto for Self-Protecting Software’, setting out the three tenets underpinning Cycode’s approach: Control, Context, and Autonomy.

��

What ADLC Security Actually Does

��

ADLC Security is built around four core capabilities. AI Visibility automatically discovers shadow AI tools, coding assistants and Model Context Protocol (MCP) servers across the development environment, giving security teams a clear picture of what AI is actually in use – including unapproved tools that exist outside standard oversight.

AI Governance enforces policy-driven control over which and models developers can use, with full AI Bill of Materials (AIBOM) coverage for compliance frameworks including SSDF, NIST, SOC2 and ISO 27001. AI Guardrails blocks risky patterns and prompt-leaking secrets in real time at the IDE and command line interface level, stopping unsafe outputs before they enter the codebase. AI Risk Detection scans application code for OWASP LLM Top 10 vulnerabilities – the class of AI-specific security weaknesses that legacy Static Application Security Testing (SAST) tools were not built to identify.

The Platform Architecture

��

Every signal from ADLC Security feeds into Cycode’s Context Intelligence Graph (CIG), a semantic, relational and temporally-aware layer that powers AI reasoning across the platform. Cycode Maestro, its agentic security orchestration engine, then handles triage, prioritisation, remediation and prevention automatically. The result is a platform that brings together AI Code Security, Software Supply Chain Security, Risk Posture Management and ADLC Security under a single graph and a single agentic engine.

Katie Norton, Research Manager at IDC, noted the significance of the launch: “Agentic development is giving rise to a new paradigm for software delivery – the Agentic Development Life Cycle – and introducing a new risk profile for enterprise security teams. As delivery becomes more automated and autonomous, organisations need security platforms purpose-built for these workflows. Solutions such as Cycode are addressing these requirements by connecting , application security controls, and remediation capabilities in a single, integrated offering.”

��

Credentials And Customer Base

��

Cycode already has a track record in enterprise application security that gives the launch context.

The company was ranked number one for Software Supply Chain Security in Gartner’s 2025 Critical Capabilities for Application Security Testing, recognised as a Leader in the 2025 IDC ASPM MarketScape and named a Leader in the 2025 Frost Radar for Application Security Posture Management across both Innovation and Growth. Its customer base includes global enterprises across , retail, manufacturing and software – including multiple Fortune 500 companies.

ADLC Security is generally available now to all Cycode customers. More information and demo requests at cycode.com.

The post Cycode Wants To Secure The Agentic Era – And It’s Just Launched The Product To Prove It appeared first on 91̽��.

OTT Cybersecurity LLC, the company behind , today announced several milestones that together position the company as foundational security infrastructure for the agentic AI era: the deployment of a real-time zero-day tracking and disclosure system designed to notify affected organisations of active exploit activity; acceptance into Anthropic’s Cyber Verification Program (CVP) and the public release of the Agent Trust Protocol (ATP), an open cryptographic standard for securing AI agents operating autonomously on the internet.

��

Real-Time Zero-Day Tracking: Finding Threats Before They Become Breaches

��

The cybersecurity industry has long struggled with a dangerous gap: vulnerabilities can be exploited in the wild for days, weeks, or months before organisations running affected systems are notified. Lyrie is designed to close that gap.

Lyrie’s autonomous threat intelligence engine continuously monitors global infrastructure, open-source repositories, API surfaces and agent-to-agent communication channels to identify zero-day vulnerabilities as they emerge. When a zero-day is confirmed, Lyrie’s system can generate a disclosure package that includes proof-of-concept analysis, impact assessment and remediation guidance for affected companies and organisations.

In verified cases, Lyrie has tracked and disclosed active vulnerabilities affecting enterprise and critical infrastructure environments across multiple sectors. Affected organisations have received patch packages and remediation guidance from Lyrie’s team within hours of discovery, not after public disclosure.

“The difference between a breach and a near-miss is usually measured in hours. We built Lyrie to be the system that finds the threat before it finds you and tells you exactly what to do about it.” said Guy Sheetrit, CEO and Founder of OTT Cybersecurity LLC, the company behind Lyrie.ai.

��

Acceptance Into Anthropic’s Cyber Verification Programme

��

OTT Cybersecurity LLC was accepted into Anthropic’s Cyber Verification Program (CVP), Anthropic’s framework for verifying legitimate dual-use cybersecurity operators. CVP acceptance supports Lyrie’s work around vulnerability research, offensive security tooling, and red-team workflows on Claude’s AI infrastructure, subject to Anthropic’s applicable safety and security policies.

“Being among the first companies accepted into Anthropic’s Cyber Verification Program validates what we’ve built. Lyrie isn’t a security tool that sits alongside AI. It’s the security layer that AI runs on top of”.

The Agent Trust Protocol: A Cryptographic Standard For AI Agent Identity

��

Enterprises and governments are deploying autonomous AI agents at unprecedented speed — agents that read email, write code, move money, sign contracts, and act on behalf of human operators. The security model for those agents has not existed at enterprise scale. Lyrie was built to change that.

The Agent Trust Protocol (ATP), authored by Lyrie’s research team and published openly at lyrie.ai/research, is a cryptographic standard that lets any system verify, in real time:

• Identity — who the AI agent is
• Scope — what it is authorised to do
• Attestation — whether it or its instructions have been tampered with
• Delegation — who delegated authority
• Revocation — whether that authority has been revoked

“Every AI agent on the internet today is a stranger. You don’t know who it is, what it’s authorised to do, or whether it’s been tampered with. ATP is the protocol that changes that.” Guy adds.

The protocol is open, royalty-free, and slated for submission to the Internet Engineering Task Force (IETF). The reference implementation is published under MIT license at github.com/OTT-Cybersecurity-LLC/lyrie-ai.

-This is a paid press release published via TechNewswire-

The post Lyrie.ai Deploys Real-Time Zero-Day Tracking Across Global Enterprise Infrastructure appeared first on 91̽��.

Security groups and technology companies spent years telling people to create longer passwords or to add symbols and avoid reusing the same codes but now, the conversation has turned toward removing passwords from daily life completely.

Steve Shoaff, SVP of Transformation at Imprivata, thinks the old system no longer makes sense for modern security.

He said, “Today is World Password Day – a reminder of one of the most outdated and frustrating conventions still embedded in modern technology. Passwords have long been a necessary part of digital security, while at the same time being one of its biggest liabilities.

“Bad password habits have been around for so long that continuing to blame users just isn’t productive. The real problem is that the model itself is broken and increasingly unnecessary for the majority of our logins.”

Shoaff believes password free logins are getting close. He said, “That’s why I’m hopeful this may be one of the last remaining World Password Days. The industry is moving toward a future where passwords fade into the background – or disappear entirely -replaced by stronger, smarter authentication methods built on cryptography, trusted devices and identity-bound access.”

��

What Is Replacing Passwords?

��

The UK’s National Cyber Security Centre announced should now become the first login choice for consumers when websites and apps offer them. The NCSC, which is a part of GCHQ, said passwords no longer offer enough protection against modern cyber crime.

Passkeys let people sign into accounts using a phone, fingerprint or facial scan instead of typing passwords manually. The NCSC says that passkeys are usually more secure than even very good and complicated passwords used with two step verification. The organisation also said most cyber crime against individuals begins when criminals steal login details.

Google data released through the NCSC showed that more than 50% of active Google users in the UK already have at least one passkey registered. Large online services including Google, eBay already support the technology.

Jonathon Ellison, Director for National Resilience at the NCSC, said, “Adopting passkeys wherever you can is a strong step towards a safer, simpler login experience and I am pleased that we can now support uptake.”

He also said, “The headaches that remembering passwords have caused us for decades no longer need to be a part of logging in where users migrate to passkeys – they are a user-friendly alternative which provide stronger overall resilience.”

Shoaff believes the biggest gain comes from removing human memory from security systems. He said, “The goal shouldn’t be better passwords or . It should be a world that no longer asks people to manage passwords at all.”

��

More Experts Answer: Is This The End Of World Password Day?

��

As more professionals recommend passkeys over passwords, the question of whether World PassWord Day is still a relevant title is being discussed. More experts weigh in…

��

Our Experts:

��

• Niall McConachie, Regional Director (UK & Ireland), Yubico
• Jeff Watkins, Chief AI Officer, NorthStar Intelligence
• Terry Lewis, Founder and CEO, RoboShadow
• Jon Kane, Senior Director, Europe & META Channel, Forcepoint
• Kamran Bahdur, Chief Information Officer, FLR Spectron

��

Niall McConachie, regional director (UK & Ireland), Yubico

��

��

“Traditional passwords are fundamentally flawed and increasingly vulnerable to compromise – a major concern given they are still the most commonly used authentication method3, leaving users highly susceptible to cyber attacks like phishing. This reality is even more alarming amid the increasing sophistication and evolution of AI-powered threats. Cyber criminals are no longer simply using AI to write phishing emails; they are deploying autonomous agents that can plan, reason and execute multi-stage attacks without human oversight.

“In response to the evolving threat landscape, users must move away from passwords towards stronger, more resilient technologies. The clear successor is the passkey, which is now the gold standard for secure, modern authentication in a digital world. This shift is gaining momentum globally and is being embraced across industries. For example, the UK Government is already in the process of adopting passkeys for its digital services, citing the superior security and protection they provide4.

“In its most secure form, a passkey is device-bound – it is not a secret that staff must remember (like a password), but a physical token they possess – such as a hardware security key. The passkey is stored on the physical device and is resistant to phishing because it cannot be intercepted or stolen by remote attackers, meaning only the key holder can gain access to their accounts. They also manage logins across all users’ platforms and devices – meaning attackers can’t use AI to get around the wall of defence the physical key provides.

“With phishing-resistant multi-factor authentication (MFA) available to all, there’s no need to continue using insufficient authentication methods like passwords to keep online accounts secure. This World Passkey Day, it’s time for the widespread use of hardware-backed passkeys to take off and for passwords to be left in the past.”
��

Jeff Watkins, Chief AI Officer, NorthStar Intelligence

��

��

“As the world moves away from the venerable password towards more secure passkeys, World Password Day may eventually become an artefact of the past: remembered fondly, perhaps, but with a sense of wonder that it was ever needed in the first place.

“For years, we put the burden of password security on end users. That may have been tolerable in a simpler digital world, but today people are expected to secure dozens of services, accounts, and devices, often while being told to create long, unique, complex passwords that they must never reuse and, ideally, never write down. That was never a realistic human-centred security model.

“Passkeys change that equation. They shift authentication away from something users must remember, manage, and repeatedly defend towards something more secure by design. In that sense, passkeys are not just a better password; they are a recognition that passwords were always asking too much of people.

“World Password Day may persist for longer than we expect, though. Security is still too often treated as a cost centre rather than a value generator, and passwords will not disappear overnight. Legacy systems, poor implementation, user habits and uneven adoption mean we are likely to live in a hybrid world for some time yet.

“Will we see a “World Passkey Day” take its place? I’m not convinced. If passkeys are implemented properly, they should feel almost invisible. We tend not to create awareness days for security controls that simply work in the background. We do not have a World SSO Day or a World MFA Day, though Global Encryption Day exists, so I would not rule it out entirely.

“Will anyone miss passwords when they are gone? I suspect not. I am already looking forward to the eventual obituary:

“The password, source of countless cybersecurity breaches, born in the early days of computing and dragged unwillingly into the modern internet age, quietly slipped away last night in the year 2045. We tried making them longer. We tried making them more complex. We tried stopping people from writing them down. We tried stopping people from reusing them. We even tried password managers. But in the end, the password simply did not fit the way humans live online. Loved by hackers, hated by everyone trying to remember whether they needed a capital letter, a number, a symbol and the name of their first pet. No flowers.”

��

Terry Lewis, Founder and CEO, RoboShadow

��

��

“World Password Day 2026 Needs a Reset to “Cyber Discipline Day”

“World Password Day made sense when passwords were the front line of security, but in 2026, that’s no longer the case.

“Today, most organisations already have access to enterprise‑grade security by default. Multifactor authentication is widely available, passkeys are native to modern devices, and hardware‑backed protections like TPM are standard. The issue, therefore, isn’t about technology; it’s about discipline, and whether organisations use it consistently.

“In the AI era, attackers aren’t manually guessing passwords. They’re using automation to continuously scan, probe and enumerate environments at scale. Whether it’s a weak credential, an exposed API key, or a forgotten device, anything visible will eventually be tested.

“The real shift is that enumeration is no longer silent can detect it.

“Modern security tooling, including SIEM and SOC capabilities, is now more accessible than ever. That means organisations can see when accounts are probed, when credentials are tested, and when unusual authentication patterns emerge, even in environments using MFA or passkeys. AI hasn’t broken security, but it dramatically increases the volume and persistence of these attempts. It creates constant background noise from systems being tested, credentials being tried and access points being explored.

“The organisations that win aren’t those with the most complex or longest password policies; they are the ones that can see this activity, understand it, and respond to it quickly.

“In 2026, security isn’t about better passwords. It’s about cyber discipline and having the everyday operational habits that keep environments clean, visible and resilient.”

��

Jon Kane, Senior Director, Europe & META Channel, Forcepoint

��

��

“When I first started in cyber, I worked with someone who had ‘Tipp-Exed’ their passcode onto the back of their 2 factor authentication device! This certainly wasn’t acceptable then but shows that security is not just about the tech; its policy and education too. The rise of social engineering threats and phishing scams – which rely on human error – are forcing users to rethink their passwords and broader security strategy.

This World Password Day, organisations need to rethink the ways they secure their networks, not just technology but also how it is used. Recent guidance from the NCSC recommends moving from passwords to passkeys and other biometric identity methods – a reflection of the changing nature of our identity security landscape.”

��

Kamran Bahdur, Chief Information Officer, FLR Spectron

��

��

“I would certainly not say World Password Day is coming to an end, as passwords will not disappear overnight just because a newer, more secure authentication method has appeared. Many systems and applications will continue to rely on them for years to come until passwords are fully phased out. This means they will still need to be maintained, and best practices regarding passwords will still need to be upheld.

“The transition to the passkeys will continue, however, and will only accelerate as become increasingly sophisticated. As much as passkeys provide a much stronger defence against traditional password-based attacks, I am sure attackers will continue to find ways around them in some form. It is even plausible we will see another layer of authentication introduced as major threat actors shift their attention away from dwindling classic combinations of passwords and MFA.

“I think World Password Day is here to stay even if the original meaning of it shifts towards newer and more secure authentication methods.”

The post Part 1: Is This The End Of World Password Day? Experts Weigh In appeared first on 91̽��.

Claude has been targeting organisations across multiple countries. Grok was deployed in national security contexts despite a documented history of generating harmful content. The Pentagon designated Anthropic a supply chain risk after it refused to remove ethical constraints on autonomous weapons use, a decision in a legal challenge that some commentators have described as likely unlawful First Amendment retaliation. Vibe-coded applications are shipping with significantly higher vulnerability rates than human-written code.

The incentive structure underneath all of this is quite consistent. When competitive position depends on how fast a product ships, safety functions get negotiated down. Internal safety teams are routinely under-resourced relative to the product and research organisations they are supposed to govern.

A 2025 EY-linked survey found that a majority of organisations allow employees to develop or deploy AI agents without high-level approval; only 60% issue formal guidance for such work. OpenAI significantly restructured its internal safety and alignment functions in early 2026. FLI’s Winter 2025 Index concluded that no frontier lab scored in the top tier on overall safety, with scores on existential-risk measures particularly weak.

The case that this represents a structural problem rather than a collection of isolated incidents is building. The attack surface created by AI systems is qualitatively different from previous software: these are systems that can autonomously take actions, adapt to context and be redirected toward purposes their developers did not intend. A model capable of identifying vulnerabilities rapidly and across large codebases can also be used to exploit them. A model trained to be helpful can be prompted to assist with extortion.

The debate has shifted: no longer whether AI has introduced new risks, but whether the industry has the infrastructure to contain them.

��

A Structural Imbalance, Not A Collection Of Mistakes

��

The pattern across the evidence points not to individual companies cutting corners, but to competitive dynamics making caution commercially irrational. Labs that move slowly lose ground and safety teams that block deployment get overruled.

Researchers who probe vulnerabilities in publicly available models face threats of legal action under terms of service that prohibit safety-related testing. In 2023, major AI labs signed White House voluntary commitments to support independent safety research. By 2024, almost none had established real protections for the researchers who try to do it.

At the , the problem is compounded by data environments that were not designed for AI. Fragmented data, inconsistent classification policies and limited visibility into where sensitive information flows mean that integrating AI systems into existing infrastructure significantly increases the risk of unintended exposure. The pace of AI adoption has not been matched by the governance maturity needed to make it safe.

The problem with safety teams isn’t competence – they are being outpaced by an architecture that treats security as a gate at the end of the process rather than a foundation at the start of it. The result is that safety functions are reviewing products that have already been shipped, patching vulnerabilities in systems already in production, and managing incidents in real time rather than preventing them at the design stage.

��

What Meaningful Safety Infrastructure Would Actually Look Like

��

There is rational agreement across the field on what the components would need to be.

Independent, adversarial safety research with genuine legal protection rather than the threat of litigation. Mandatory pre-deployment testing with enforcement teeth rather than voluntary frameworks. Zero-trust deployment environments where AI agents operate under least-privilege constraints and require cryptographic human-in-the-loop authorisation for sensitive actions. AIBOM manifests bound to runtime telemetry. Incident disclosure requirements that create accountability for failures rather than allowing them to be buried.

The real challenge lies in identifying who will build this. Labs are competing against each other and have weak incentives to absorb the cost of infrastructure that benefits the whole industry. Enterprise buyers could in principle refuse to purchase models that lack transparent governance, but most currently lack the technical authority to audit what they are buying.

Regulators have the mandate but have consistently lagged the technology. The EU AI Act is the closest thing to a binding framework; the US has no real equivalent. What several contributors to this piece argue is that until the cost of deploying an insecure AI system exceeds the commercial benefit of having shipped it first, will remain just that.

We put the question to AI safety researchers, cybersecurity specialists and deployment risk experts to find out what they think needs to change.

Our Experts:

��

• Omair Manzoor, Founder and CEO, ioSENTRIX
• Paulo Cardoso do Amaral, former CIO and NATO Scientific Advisor on Cybersecurity
• Raphael Karger, CTO, ZeroPath
• Seb de Lemos, CEO, hosting.com
• Shreyans Mehta, CTO, Cequence Security
• Collin Hogue-Spears, Senior Director, Black Duck Software
• Stanislav Kazanov, Head of GRC, Cybersecurity and Sustainability, Innowise
• Aviral Srivastava, Security Engineer, Amazon

��

Omair Manzoor, Founder and CEO, ioSENTRIX

��

“The honest answer is yes, but not in the way most people frame it. The problem isn’t that any single company decided to cut corners. It’s that the competitive dynamics made cutting corners rational. When the competitive gap between shipping now and shipping in six weeks determines market position, safety stops being a foundation and becomes a negotiable variable.

“We’re seeing the results in real time. Claude Code weaponised into an automated extortion pipeline. Apple Intelligence hijacked through prompt injection on 200 million devices. shipping with three times the vulnerability rate of human-written code. These aren’t hypotheticals. These are findings from our actual pen testing engagements and from public research in the last few months alone.

“Safety teams can’t keep pace, not at current resourcing levels. The product team ships the LLM integration before the security team knows it exists. Shadow AI is the new shadow IT, except it moves faster and touches far more sensitive data. What meaningful safety infrastructure looks like is honestly pretty boring: mandatory adversarial testing before any model touches production data, independent red teaming that isn’t funded by the company being tested, and regulatory teeth. Not guidelines, not frameworks. Actual enforceable standards with consequences. Until the incentive structure rewards caution, we’ll keep having this conversation every time something blows up.”

��

Paulo Cardoso do Amaral, Former CIO and NATO Scientific Advisor on Cybersecurity

��

“The AI race has structurally compromised safety, not because every model is reckless, but because the incentives are. When speed, scale and strategic positioning dominate, safety becomes a drag coefficient rather than a hard launch condition. Attackers can automate code exploitation faster. Social engineering is now powered by convincing voice, image and video impersonation. Frontier models are being pulled into national security contexts before governance is mature.

“Safety teams are not keeping pace. In too many organisations, advisory functions remain while product and deployment teams operate at wartime tempo.

“Meaningful safety infrastructure would look more like aviation or : mandatory pre-deployment testing, independent red-teaming, continuous monitoring, incident disclosure, auditable logs, strong identity and provenance controls, and clear restrictions for military and other high-risk uses. It also requires redesigning insecure digital architectures, not merely adding guardrails afterwards. Responsibility starts with frontier labs, but deployers, regulators, sector bodies and states all share it. If AI is now part of critical infrastructure, safety cannot be a voluntary culture. It has to be engineered, audited and enforced.”

��

Raphael Karger, CTO, ZeroPath

��

“Yes, but it’s more precise to say the AI race has revealed a pre-existing structural gap. Security has always been an afterthought in software. AI just accelerated the timeline and raised the blast radius. The pressure to ship isn’t new. What’s new is that the models being shipped can themselves be weaponised as attack infrastructure. The race dynamic makes it harder to justify slowing down for security work that doesn’t show up on a benchmark.

“Safety and security teams at most AI labs are structurally downstream of the product and research organisations. They review what’s already been built. That’s not a staffing problem. It’s an architectural one. You can’t hire your way out of a process that treats security as a gate rather than a foundation.

“Meaningful infrastructure means continuous, automated security validation integrated into the model development lifecycle, not red-teaming sprints before a release. It means treating AI systems like the complex attack surfaces they are. Responsibility is shared: labs own the model layer, but the broader ecosystem, the platforms, integrations and deployment environments, needs its own security posture. Right now almost no one is looking at that layer seriously.”

��

Seb de Lemos, CEO, hosting.com

��

“AI hasn’t broken safety outright overnight, but it has materially stretched and fragmented it, particularly in software development. With AI, anyone can now act as a developer. That democratisation is powerful, but it introduces uneven standards, where production-ready code is deployed without the governance, testing and review processes that were once standard. Many people developing software now either don’t fully understand what they’re building or are using AI to accelerate development without understanding what loopholes their code might contain.

“Internal security teams are being asked to operate at a pace and scale that simply didn’t exist before. AI accelerates development, but security practices, governance processes and compliance checks have not scaled at the same rate. We’re seeing this play out in real incidents where AI-generated code has introduced vulnerabilities because the underlying logic wasn’t validated. Safety teams aren’t failing. They’re being outpaced.

“Meaningful safety infrastructure needs to be built in, not bolted on, spanning the full lifecycle from development through to deployment and ongoing maintenance. should be operationalised directly into infrastructure, ensuring applications are compliant by default rather than through manual intervention. If AI is lowering the barrier to building software, the industry must equally lower the barrier to building it safely.”

��

Shreyans Mehta, CTO, Cequence Security

��

“The cybersecurity industry spent a decade building detection around human behavioural signals. AI agents break that detection. They make direct HTTP requests from clean residential IPs with plausible headers, never execute JavaScript, never render a page. Every UEBA baseline built on human behavioural norms is now effectively irrelevant. What matters now is real-time detection: server-side behavioural analysis trained on years of real API traffic, operating on mathematical models that do not depend on the entity being human.

“Most organisations that have moved beyond basic connectivity have landed on identity as their answer. Integrate with an enterprise identity provider, enforce OAuth, and ensure agents act on behalf of authenticated users. But this is exactly where the industry’s thinking stops, and where the most dangerous failures begin. Controlling agent permissions at the tool level is essential, not just who the agent is, but what it is allowed to do.

“Sensitive data still flows through tool calls that identity alone cannot inspect. Agent behaviour can drift in ways that authentication cannot detect. This is why AI gateways are needed: combining sensitive data detection, behavioural fingerprinting, session binding and a trusted registry on top of identity and connectivity. One AI coding agent we observed made 2,500 tool calls over 48 hours before improvising, probing unauthorised file paths and attempting write operations its credentials did not permit.”

��

Collin Hogue-Spears, Senior Director, Black Duck Software

��

“Yes. The EU and China have binding regulatory floors. The US does not. The December 2025 White House executive order pre-empts state action without replacing it, leaving California’s SB 53 and New York’s RAISE Act as the de facto national standard. FLI’s Winter 2025 Index graded no frontier lab above C+ overall on safety, and none above D on existential safety. The February 2026 Pentagon supply chain designation punished Anthropic, the lab with the highest safety score, for holding two narrow ethical red lines. That is the signal every other lab reads.

“Safety teams can’t keep pace, and the reason is architectural. Deterministic compliance frameworks cannot govern stochastic agents generating novel outputs on every invocation. CrowdStrike’s 2026 Threat Report puts adversary breakout time at 27 seconds. Non-human agent identities now outnumber human identities 82 to one, and only 18% of security leaders trust legacy identity access management for those agents. OpenAI dissolved its Mission Alignment team in February 2026. This is not an effort problem. It is a tool-category problem.

“Meaningful infrastructure requires an agent zero-trust gateway applying NIST SP 800-207 to every tool invocation, with deny-by-default access and scoped credentials per action; AIBOM manifests bound to runtime telemetry alerting on out-of-manifest calls; capability-tiered controls; and a pre-deployment testing framework covering prompt injection and tool misuse. NIST owns the AI RMF, OSCAL and SBOM ecosystems. That’s where the baseline gets built.”

��

Stanislav Kazanov, Head of GRC, Cybersecurity and Sustainability, Innowise

��

“The AI race has actively penalised safety. When a government blacklists a lab building frontier AI for enforcing responsible development ethics on autonomous weapons, while promoting developers who remove those constraints, the market receives a very clear signal: caution in AI development is a commercial liability.

“Safety teams are mathematically challenged to keep up. They are attempting to defend against exponential increases in capability with only linear resources. Attackers are already using vibe-hacking to exploit agentic AI tools for automated data extraction and extortion. A corporate red team cannot manually patch behavioural vulnerabilities faster than the underlying model can generate new, unforeseeable logical paths.

“Meaningful safety infrastructure in 2026 cannot consist of an internal trust and safety committee reporting to a Chief Revenue Officer. There must be zero-trust deployment environments where autonomous AI is denied from conducting privileged network functions without a hardware-bound, human-in-the-loop cryptographic signature. AI vendors cannot build this because they’re competing on price. It must be created by enterprise buyers, CISOs and GRC leaders, who will refuse to purchase models without transparent governance mandated by regulation with the technical authority to audit model weights before deployment. Until the cost of delivering an insecure AI exceeds the benefit of shipping first, the industry can only manage the blast radius, not prevent it.”

��

Aviral Srivastava, Security Engineer, Amazon

��

“The race has structurally compromised safety, but not in the way most people talk about it. The bigger risk is not rogue models. It’s that the infrastructure layer underneath these models is being shipped at startup speed with enterprise-grade security assumptions that simply are not true. I’ve filed critical vulnerabilities in AI platforms with tens of thousands of production deployments where the maintainers denied the issue, hid behind documentation as a fix, or simply stopped responding. That’s not a model alignment problem. It’s a basic software security problem dressed up in AI branding.

“Safety teams can’t keep pace because the scope of what AI safety means keeps expanding while the investment stays narrow. Most attention goes to alignment research and red teaming model outputs. Almost nobody is looking at the deployment stack, the orchestration frameworks, the model file formats, the inference engines. That’s where the actual attack surface is right now, and it’s largely unguarded.

“Meaningful safety infrastructure starts with treating AI tooling like critical software, not hackathon projects. That means funded security audits, real vulnerability disclosure programmes with actual response timelines, and regulatory teeth behind frameworks like the NIST AI RMF instead of voluntary adoption that nobody enforces. The responsibility sits with the companies shipping these tools, but most of them are currently optimising for stars and funding rounds, not security posture.”

��

For any questions, comments or features, .

��

The post Experts Comment: Has The AI Race Made The World Less Safe? appeared first on 91̽��.

According to reporting from The Register, the hacking collective behind some of the most of recent years has claimed a successful attack on the studio behind Grand Theft Auto, with a ransom deadline set for 14 April 2026.

The attackers didn’t break into Rockstar’s own infrastructure to do it. They walked through a door left open by Anodot, a third-party cloud analytics vendor Rockstar used for monitoring cloud costs and performance data.

Rockstar has confirmed that a limited amount of non-material company information was accessed and that player data and live game services were unaffected. It hasn’t disclosed the exact nature of what was taken or whether any ransom demand has been met. What it has confirmed, however, is enough to make the point: one of the best-resourced studios in gaming had its data compromised because an attacker found a weaker door in the supply chain and used it.

That detail is the one every founder building on cloud services and third-party software stacks should contemplate.

��

Same Group, Same Playbook, Different Target

��

ShinyHunters has been making headlines for years, and the pattern is consistent.

The group has previously been linked to notable breaches at Ticketmaster, Santander and AT&T, among others. A recurring characteristic of their attacks is exploiting third-party cloud infrastructure rather than attacking targets directly. , which exposed around 900,000 records, also involved ShinyHunters targeting a vendor relationship rather than the company’s core systems.

In the Rockstar case, the entry point was Anodot, a platform used for cloud analytics and cost monitoring. This bears relevance for a specific reason: analytics and monitoring tools are often given wide read access to cloud environments precisely because they need visibility across multiple systems to do their job – that makes them a valuable target. Compromise the analytics layer and you can potentially see a great deal of what the company sees, without ever touching the production systems themselves.

Rockstar isn’t the only business exposed to this. A 2021 PwC survey found that only around 40% of organisations comprehensively assess third-party and supply chain cyber risk, despite this being one of the most consistently exploited attack vectors.

The attack surface created by SaaS tools and cloud integrations is often larger than the one created by a company’s own code, and it receives far less scrutiny.

��

Your Vendors Are Part Of Your Attack Surface Whether You Like It Or Not

��

Third-party tools exist in a mental category, somewhere between ‘vendor problem’ and ‘not our responsibility’. Most startups think about security in terms of their own stack: their application code, their database, their infrastructure. The Rockstar breach illustrates why that framing is wrong.

When you give a SaaS vendor access to your cloud environment, even read-only access for monitoring purposes, you’re extending your security perimeter to include theirs. Their misconfiguration becomes your breach, their compromised credentials become your incident, and their incident response plan, or lack of one, becomes your problem to manage, including your obligations under UK GDPR and, for businesses operating in the EU, the NIS-2 Directive.

The challenge is that modern startups typically have dozens of third-party integrations touching their systems – analytics platforms, CRMs, payment processors, logging tools, CI/CD pipelines, identity providers – each one is a potential entry point. Most are assessed at onboarding and then rarely revisited.

‘we checked them at onboarding’ and ‘we haven’t looked since’ is exactly where attackers look for opportunities.

��

The Practical Bit For Founders

��

Here are three areas to address before an incident makes them urgent.

The first is vendor inventory and access review. Maintain an up-to-date list of every third party with access to your data or systems, what level of access they have and when that access was last reviewed. Any vendor with broad cloud access, like an analytics platform or a monitoring tool, warrants particular scrutiny. Ask for evidence of SOC 2 or ISO 27001 certification, penetration test summaries and incident response procedures. If a vendor can’t provide these, that tells you something important.

The second is limiting what any single integration can see or do. Cloud-native controls, strict IAM policies, VPC segmentation and isolated build environments all help limit how far the damage can spread if a third-party tool is compromised. The goal is to ensure that if vendors fail, the damage stays contained rather than cascading across the rest of your infrastructure.

The third is making sure your incident response plan explicitly covers third-party-led breaches. Many incident response plans focus on direct attacks against company systems. Rockstar’s situation is a reminder that you can be breached even when your own infrastructure is secure. Your plan should cover how you’d detect a vendor-side compromise, how you’d respond, and what your legal and communications obligations are under UK GDPR when the breach originates outside your own systems.

��

Don’t Pay The Ransom

��

On the ransom demand specifically: law enforcement and security authorities consistently advise against paying. Payment neither guarantees the deletion of stolen data nor prevents the group from attacking again or selling the data regardless.

ShinyHunters has a history of multiple monetisation attempts on the same data .The calculus for is clear: engage your incident response team, notify the ICO within 72 hours where required under UK GDPR, and work with cybersecurity specialists, not the attackers.

The Rockstar breach is a reminder, although a frustrating one, that supply chain security is no longer a problem only large enterprises need to worry about. Attackers follow the path of least resistance. For growing businesses building on complex third-party stacks, that path largely runs through the tools they’ve integrated and forgotten about rather than the systems they actively protect.

The post ShinyHunters Just Hacked Rockstar Through A Supplier – Every Business Using Third-Party Software Should Pay Attention appeared first on 91̽��.

But as speed increases, so do concerns. If developers are generating code they don’t fully understand, skipping manual reviews and relying on AI-suggested dependencies, is introducing a new wave of security risks? Are we compromising quality for quantity?

The question for startups isn’t just whether vibe coding works. Rather, the question is whether it’s safe enough for real-world use.

��

The Speed Advantage Versus the Security Trade Off

��

Vibe coding dramatically lowers the barrier to entry in way we’ve never seen before. With nothing more than a few prompts, developers can generate authentication systems, databases, APIs and front-end interfaces that previously required incredibly experienced experts. For startups, that means faster MVPs, lower costs and less reliance on large engineering teams. In theory, an absolute win.

But, security experts warn that this speed often comes at the expense of proper safeguards. AI-generated code may appear flashy and functional on the surface, but it can include insecure defaults, weak validation or outdated dependencies. When developers copy, paste and deploy without fully understanding the logic, vulnerabilities can slip into production unnoticed, and these vulnerabilities can be incredibly problematic.

In many cases, vibe-coded applications are also built without traditional development processes like threat modelling, security reviews or penetration testing��– steps that normally catch problems before release.

��

��

Common Security Risks In Vibe-Coded Apps

��

One of the biggest concerns is authentication. AI tools can generate login systems quickly, but these may lack protections like rate limiting, proper session handling or multi-factor authentication. This leaves applications vulnerable to brute-force attacks or account takeovers.

Another issue is exposed secrets. Developers sometimes include API keys, tokens or database credentials directly in prompts. These values can then appear in generated code, logs or version control systems, creating serious security exposure.

Another significant issue is that dependency risks are also growing. AI tools frequently pull in libraries automatically, and developers may not check whether those packages are maintained, secure or even necessary. This can introduce supply chain vulnerabilities without anyone noticing.

There’s also the problem of over-permissioned systems. often use broad access controls simply because they are easier to implement. Indeed, without careful review, this can allow users to access data or functions they shouldn’t.

Finally, there’s the human factor – often the most significant risk. Vibe coding encourages experimentation and rapid iteration, which is great for innovation but highly risky when code moves straight from prompt to production.

��

Why Are Startups Particularly Exposed?

��

Understandably, startups are especially likely to embrace vibe coding because of what it has to offer. Smaller teams, tighter budgets and pressure to move fast make AI-generated development appealing. But, these same factors that make vibe coding so attractive also mean that security can become an afterthought.

Unlike larger organisations, startups may not have dedicated security engineers or formal review processes. That increases the risk of vulnerabilities making it into live products, especially when founders are focused on product-market fit rather than infrastructure hardening.

Of course, another big thing to contemplate here is reputational risk. A security breach early in a startup’s lifecycle can damage trust with customers and investors, and in some cases, stall growth entirely. In many cases, it may, in fact, be the en dof the road for many startups.

��

But That Doesn’t Mean Vibe Coding Isn’t Usable

��

Despite the risks, vibe coding isn’t inherently unsafe – that’s not what we’re saying. In fact, many experts argue that the real issue isn’t AI-generated code itself, but how it’s used. When treated as a starting point rather than a finished product, . It’s not a fix-all, complete solution, and it shouldn’t be used for instant gratification.

Indeed, the key is to introduce safeguards. Human review remains critical, particularly for authentication, data handling and permissions. Automated scanning tools can also help detect vulnerabilities, exposed secrets and risky dependencies before deployment.

Another common recommendation is separating prototype and production workflows. Vibe coding can be used to build MVPs quickly, but code should be refactored and hardened before going live.

Startups should also adopt basic security hygiene, including environment variables for secrets, dependency auditing, input validation and proper access controls. These steps don’t remove the speed advantage but significantly reduce risk.

��

Tossing Up Speed and Security

��

– it’s just too effective and useful – nor should it. If anything, it’s becoming a core part of modern development workflows, especially for startups trying to move quickly. The bigger question is whether teams can balance speed with responsibility and use the technology effectively and safely.

Used carelessly, vibe coding could introduce a new generation of vulnerable applications. Used thoughtfully, it could democratise software development without sacrificing security.

For startups embracing AI-generated development, the safest approach may be simple: move fast, but don’t skip the security review.

The post Is Vibe Coding Safe Or A Cybersecurity Disaster Waiting To Happen? appeared first on 91̽��.