AI agents do more than chatbots. They can access data, use tools, and complete tasks on their own, which makes them powerful but risky.

OWASP Top 10 Provides Guidance
The OWASP Top 10 for Agentic Applications lists the main risks, offers threat models, and gives practical mitigation steps.

It helps CISOs explain threats to business teams and improve security directly.

Risks Hidden from IT Teams
Many AI agents are already running in organizations without the knowledge of security or IT teams.

This increases the chance of mistakes, data exposure, or misuse.

Educating Stakeholders is Key
Security leaders often struggle to explain AI risks to business teams.

The OWASP list is designed to be easy to use and understand, helping teams make informed decisions.

Practical Focus on Real Threats
The guidance focuses on real-world, data-driven risks rather than purely theoretical concerns, giving CISOs actionable tools for threat modeling and policy decisions.

CISOs and CSOs must take agentic AI seriously and use structured frameworks to reduce enterprise risk before incidents occur.

Autonomous AI agents are being deployed rapidly across enterprises, but most organizations are not using proper identity protections.

This creates a potential security gap that could be exploited by attackers.

Why Identity Matters
AI agents communicate with each other to complete tasks.

Without strong authentication, a rogue agent could issue harmful instructions, and legitimate agents might unknowingly follow them.

Challenges in Containment
Current systems often detect misbehavior only after it happens.

Even revoking credentials may not stop the cascade of actions already triggered by a compromised agent.

Need for New Security Measures
Experts emphasize tracking every agent interaction and automating alerts to ensure that compromised instructions are nullified.

However, solutions to manage this level of complexity are not yet widely available.

Security teams must rethink identity management and control measures for agentic AI before these systems scale further.

DXS International, a UK tech provider for the NHS, reported unauthorized access to its office servers.

The company says its clinical services stayed operational and there is no confirmation that patient data was affected.

Scope and Impact
DXS supports around 10% of NHS referrals and provides tools for clinical guidance, though it does not hold central patient records.

Investigations are ongoing with NHS cybersecurity teams and external specialists to determine the full impact.

Rising Concerns Over Health IT Supply Chains
Previous attacks on UK healthcare suppliers, like Synnovis and Advanced, show that breaches of third-party systems can disrupt patient care.

Even without direct access to core records, IT interruptions have delayed treatments and required staff to revert to manual processes.

Regulatory Context
Current UK regulations do not automatically impose strict cybersecurity standards on third-party providers like DXS.

The government’s new Cyber Security and Resilience Bill aim to extend obligations to critical service IT suppliers, with fines for noncompliance.

Securing health IT providers is crucial, as third-party breaches can indirectly threaten patient care and operational continuity.

La Poste, France's national postal service, experienced a massive network incident that took down digital platforms, including its website, mobile app, and document storage system.

Impact on Services
Although some post office locations faced service delays, customers could still complete banking and postal tasks at counters.

Online payments with SMS authentication and ATM withdrawals remained functional.

Banking Arm Operations
La Banque Postale confirmed that core banking processes, such as interbank transfers and cash withdrawals, continued to operate despite the outage.

Mobile and online card payments were temporarily adjusted using SMS codes.

Cause and Context
Reports suggest the disruption was caused by a distributed denial-of-service (DDoS) attack affecting operations nationwide.

La Poste employs over 250,000 staff and provides a range of services, from mail delivery to banking and telecommunications.

Cyberattacks on critical national services demonstrate the need for robust resilience planning and rapid incident response to minimize disruption to essential operations.

Malware on a Ferry Raises National Security Concerns

Arrests and Charges
French authorities detained a Latvian crew member aboard the Italian ferry Fantastic, part of Grandi Navi Veloci.

He faces charges of conspiring to access computer systems on behalf of a foreign power. A Bulgarian suspect was released without charges.

How the Malware Was Found
The ferry's operator detected suspicious software while docked in Sète.

The malware could have allowed remote control of the ship, but it was neutralized without causing harm.

French and Italian authorities are jointly investigating the incident.

National Security Implications
French Interior Minister Laurent Nuñez confirmed suspected foreign interference.

The investigation is being conducted alongside counterespionage units, highlighting risks of cyberattacks targeting critical transportation infrastructure.

Ongoing Legal Process
The Latvian suspect has been transferred to Paris for legal proceedings. French authorities continue to examine seized items for evidence and are assessing broader security impacts.

This case underscores the growing risk of cyberattacks on essential transportation systems and the need for heightened vigilance.

AI is becoming a useful tool in cybersecurity, helping teams process huge amounts of data quickly and spot hidden threats.

Balancing Speed and Certainty
AI can detect anomalies and prioritize risks far faster than humans, but it cannot make final security decisions on its own.

Deterministic systems must handle actions like blocking accounts or granting access to ensure consistent, auditable results.

Limitations of AI Alone
AI outputs can vary over time, models can drift, and attackers can manipulate inputs.

These factors make AI unreliable as a sole guardian for security enforcement.

Compliance and audit requirements also demand transparent, reproducible decision-making.

Human Oversight Remains Critical
Security teams should use AI to inform decisions, not replace rules or human judgment.

Even when AI provides recommendations, humans must validate actions where errors could impact systems or data.

Testing for Determinism
Before applying AI, organizations should ask six questions about audit expectations, risk of mistakes, and potential manipulation.

If any answers are yes, deterministic controls must remain in place.

AI can speed analysis and suggest actions, but humans and deterministic systems must retain ultimate control to protect critical systems and data.