Why the UK’s Cyber Spies Are Worried About AI Agents

The UK’s National Cyber Security Centre has a message for every organization building autonomous AI agents: assume they’ll misbehave. The agency published interim practical advice on August 20 for firms designing or operating agentic systems, and the tone is unmistakably cautious.

Several incidents involving AI models carrying out unsanctioned or unintended actions prompted the guidance. The NCSC admits formal guidance is still in the works and will eventually supersede this blog post. For now, this is the blueprint.

The advice builds on earlier NCSC work on securing agentic AI and lands as organizations scramble to build governance frameworks for systems that increasingly act on their own.

Sandboxing: The First Line of Defense

The NCSC’s core recommendation is refreshingly simple: figure out how much autonomy your agent actually needs before you let it loose. The agency wants firms to threat-model the agent’s prompts, tools, networks, and accessible services, then use those findings to decide which extra controls are necessary.

Don’t rely on the safeguards baked into the underlying model or agent framework. The NCSC warns these can be bypassed or prove insufficient in higher-risk environments. That’s a pointed message for teams who assume the AI vendor has it covered.

Network Controls That Say “No” by Default

For higher-risk deployments, the NCSC recommends running agents in robust AI sandboxing techniques and restricting access to only the resources required for a task. Network controls should deny connectivity by default wherever possible, with allowlists or service-aware proxies for anything the agent genuinely needs to reach.

The agency also advises separating agent execution, supporting infrastructure, and inference services where feasible. Why? Because agents can potentially discover configuration weaknesses or vulnerabilities in their own technical controls, creating a real risk of sandbox escape. That’s not paranoia — it’s a known failure mode.

Identity and Credentials: Shrink the Blast Radius

Every agent should get a distinct identity, the NCSC says, with credentials limited to what the task requires. Short-lived credentials are preferred. And here’s the kicker: organizations should treat API keys, OAuth grants, SSH keys, and authenticated sessions as part of an agent’s potential “blast radius.”

That framing is useful. If an agent goes rogue, those credentials are the difference between a contained incident and a full-blown breach. The less access each agent has, the smaller the damage it can do.

Human Oversight: Not Optional

The NCSC wants humans in the loop for higher-risk activity. That means named responsibility for agent operations, real-time monitoring, and the ability to intervene when unexpected behavior occurs. No more letting the AI run unsupervised and hoping for the best.

Agent activity should be logged and monitored as part of security operations and incident response. The agency also stresses the importance of being able to halt autonomous activity immediately — including restricting network access and communications with model infrastructure when necessary. A kill switch, essentially.

This aligns with broader industry efforts. The OWASP agentic AI security framework introduced earlier this year takes a similar maturity-based approach to managing these risks.

What This Means for Your AI Strategy

The guidance isn’t meant to scare organizations away from agentic AI. It’s meant to make them think before they deploy. The NCSC explicitly says the advice should evolve alongside the technology, and that organizations should regularly reassess whether the autonomy granted to agents remains proportionate to their risk tolerance.

That’s the key takeaway: autonomy is a privilege, not a default. Start with less, monitor closely, and scale up only when you understand the risks.

For teams building agentic systems, the checklist looks like this:

• Assess required autonomy and threat-model before deployment
• Run higher-risk agents in sandboxes with minimal resource access
• Deny network connectivity by default; use allowlists
• Give each agent a distinct identity with short-lived credentials
• Maintain human oversight with real-time monitoring and intervention capability
• Log everything and ensure you can halt activity instantly

The NCSC’s interim advice is a wake-up call. Autonomous AI is powerful, but it’s also unpredictable. Organizations that treat security as an afterthought will learn that the hard way. Those that follow this guidance have a fighting chance.

For more on securing AI systems, check out our coverage of AI governance best practices and autonomous agent risk management.

UK Fraud Cases Hit Record High in 2026: The Numbers

Fraud in the UK has never been this rampant. Over 220,000 cases were filed with the National Fraud Database (NFD) between January and June — the highest number ever recorded in the first half of a year, according to Cifas. That’s a sobering statistic for consumers and businesses alike.

The non-profit, which runs the NFD and the Insider Threat Database, reports that identity fraud alone rose 9% year-on-year to nearly 130,000 cases. This surge is largely driven by scammers targeting bank accounts and plastic cards, which account for 68% of all identity fraud cases. Impersonation incidents using the victim’s real address also jumped 12% YoY.

While “false identity” filings dropped 35% YoY, mainly in banking and telecoms, the report warns that “intelligence continues to indicate growing concerns around synthetic identities, AI-enabled impersonation and digitally manipulated documentation.”

Account Takeover and SIM Swap Fraud on the Rise

Account takeover cases are another major driver of UK fraud in 2026. Cifas recorded nearly 40,000 such cases, a 5% YoY increase. Online retail incidents soared 84%, and card account cases rose 59%. But the most alarming trend is unauthorized SIM-swap fraud, which skyrocketed 402% to 4,109 incidents — now representing 10% of all filings, up from just 2% a year ago.

This explosion in SIM swapping highlights how criminals exploit mobile networks to bypass two-factor authentication and drain accounts. If you’re not aware of this tactic, it’s worth understanding: fraudsters convince mobile carriers to transfer your number to a SIM they control, then use it to reset passwords and access your financial accounts.

Young Adults: Both Victims and Perpetrators

Identity fraud victims are getting younger. While the 61-and-over age group still accounts for the most cases, the biggest increase came in the 21-30 age bracket, where cases rose by almost a third (32%).

But here’s the twist: people under 30 also represent a majority (57%) of money muling cases, with 17% under 21. Money muling — where individuals let their bank accounts be used to transfer stolen funds — increased 69% annually, with over 13,000 filings. Mule activity now accounts for 30% of all cases of misuse of facility.

Part of this increase stems from better detection and a new filing reason introduced in 2025: “funds received – money muling.” Still, the figures suggest a worrying trend of young people being recruited into fraud networks, often through social media “job ads” that promise easy money.

Why Is Money Muling Growing?

Cifas CEO Mike Haley points out that identity fraud now accounts for three-fifths of all NFD cases, highlighting the value of personal information to scammers. “Whether it is used to open accounts, take over existing facilities or support wider criminal activity, stolen personal data often provides the entry point,” he said.

He also stressed the importance of early intervention: “As criminals continue to evolve their tactics and use digital channels to reach new audiences, education, awareness and prevention remain — particularly for younger people who are increasingly exposed to fraud risks.”

What This Means for Consumers and Businesses

The takeaway is clear: UK fraud is not slowing down. If you’re a consumer, protect your personal data like it’s gold — because to fraudsters, it is. Use unique passwords, enable multi-factor authentication (but be wary of SIM-swap risks), and monitor your bank statements regularly.

For businesses, the rise in account takeover and identity fraud means investing in robust verification systems is no longer optional. Consider biometric checks, device fingerprinting, and real-time fraud monitoring. For more on protecting yourself, check out our guide on how to prevent identity theft and tips for spotting phishing scams.

Looking Ahead: The Future of Fraud Prevention

Cifas’s data paints a grim picture, but it also underscores the importance of vigilance. As AI-enabled impersonation and synthetic identities become more sophisticated, both consumers and institutions must adapt. The record numbers in 2026 are a wake-up call — fraud prevention can’t be an afterthought anymore.

Stay informed, stay skeptical, and remember: if an offer seems too good to be true, it probably is. The fight against fraud starts with awareness.

Record-High Scores, Record-Low Trust

Defense contractors are reporting their best-ever cybersecurity scores under the Pentagon’s CMMC program. They just don’t believe the numbers themselves.

The CyberSheath 2026 State of the DIB Report, released August 20, found the average Supplier Performance Risk System (SPRS) score jumped to +51 — a five-year high, up from +33 in 2025. That was already the first positive score in the report’s history.

SPRS is the self-assessment tool defense contractors use to gauge their cybersecurity maturity under the Cybersecurity Maturity Model Certification (CMMC), the DoD program that governs how companies protect federal contract information and controlled unclassified information.

But here’s the catch: confidence in those scores cratered. Only 65% of contractors said they were extremely or very confident their score was accurate. A year ago, that figure was 89%. In 2024, it was 94%.

That’s a 24-point drop — and it’s the starkest finding in the entire report.

Why Contractors Doubt Their Own SPRS Scores

David M. Schneer, CEO of Merrill Research, which conducted the survey of 302 defense contractors, called the disconnect between rising scores and falling confidence “the most striking finding this year.”

“Contractors are reporting higher SPRS scores and greater adoption of important cybersecurity capabilities, but confidence in the accuracy of those scores has fallen substantially,” Schneer said. “That tension suggests that measuring progress requires looking beyond the reported score itself.”

The timing isn’t a coincidence. In July 2026, the Trump administration suspended CMMC Phase II, which would have introduced independent verification by Certified Third-Party Assessment Organizations (C3PAOs). Originally set to take effect November 10, 2026, that phase is now on hold.

Without external checks, contractors are left to grade their own homework — and they know it.

Only 1% Feel Fully Prepared

Just 1% of contractors believe they’re completely ready for CMMC certification. That number hasn’t budged since a previous CyberSheath study in October 2025.

So while scores are climbing, actual readiness isn’t. The two metrics are telling different stories.

Money Isn’t the Main Problem

The bottleneck isn’t budget — at least not in the way you’d expect.

• 53% of respondents said their cybersecurity budgets felt “just right”
• 24% said they had more than enough
• Average DFARS compliance spending rose to $155,204 annually

DFARS, the Defense Federal Acquisition Regulation Supplement, is the rulebook that makes CMMC a binding legal requirement for DoD contracts. Contractors assess themselves against 110 security controls from NIST SP 800-171, with a perfect score of 110.

The CyberSheath report argues the real challenge isn’t how much contractors spend, but “how effectively those investments translate into implemented, sustainable and verifiable security.”

Contractors Want Reform — But Not Deregulation

Here’s the twist: even as they struggle with compliance, contractors aren’t asking to ditch the rules.

90% of respondents still support a legal mandate for minimum cybersecurity standards. 77% said DFARS compliance meaningfully improves national security.

What they want is a better path to get there:

• 74% want easier implementation processes
• 70% want more vendor options to support compliance

And 52% fear losing contracts due to non-compliance. That fear is real, but it hasn’t soured them on the program’s purpose.

What’s Next for CMMC?

Emil Sayegh, CEO of CyberSheath, points out that most DIB contractors are manufacturers, engineers, and specialized businesses “whose mission is supporting the warfighter, not becoming cybersecurity experts.”

His ask to the federal administration: reform CMMC to “make effective cybersecurity easier to consume while preserving objective, verifiable assurance that the protections are actually in place and working.”

The report’s bottom line is a warning wrapped in a suggestion. Self-reported scores are rising, but if contractors themselves don’t trust them, the numbers are close to meaningless. When CMMC Phase II eventually resumes — and most observers think it will — the C3PAO verification process will be the real test.

Until then, the DIB is flying on instruments it knows are miscalibrated.