The European Commission said on 19 November that it had been targeted by an AI-generated cybercampaign. The Associated Press reported that the incident was first detected on 22 October; Commission spokesperson Thomas Regnier said several applications were temporarily taken down and most had been restored, with no evidence of a data breach affecting Commission staff as a whole.
Insurance Journal, reporting Bloomberg coverage, said the Commission was investigating any potential employee data breach and that the attack involved misuse of an artificial intelligence tool. Infosecurity Magazine reported that the European Data Protection Supervisor had been notified and that the Commission was in contact with CERT-EU.
For boards, the governance point is not only cyber. If AI tools help attackers find vulnerabilities and automate campaigns, AI deployments need logging, access controls, fallback processes and incident escalation routes that have been tested before systems become business-critical.
The FCA launched its AI Spotlight Observatory on 19 November. The regulator said the platform is based on practical use cases from more than 200 organisations and spans 150 areas of activity, with examples from financial crime and compliance to agentic payments and neurosymbolic AI.
The FCA said the Observatory adds a common taxonomy, separates use cases from implementations and will continue to add new use cases. Registration is needed for more detailed insight, and firms can apply to be showcased.
For financial-services leaders, this is a practical governance prompt. A firm's AI inventory should be granular enough to compare against the FCA taxonomy, show who owns each use case, and identify where customers, money, regulatory duties or market integrity are involved.
The Department for Education and Department of Health and Social Care published non-statutory guidance on 20 November for the safe, responsible and appropriate use of AI in children's and adult social care in England. The guidance says AI should strengthen, not replace, human connection, professional judgement and accountability.
DfE and DHSC set out four principles for social-care AI use. Practitioners remain responsible for casework and decisions; AI should not make decisions that directly affect people; AI should support better experiences and outcomes; and personal information should not be entered into publicly accessible AI tools.
Although the guidance is sector-specific, it is useful beyond social care because it shows how public bodies are treating high-consequence AI. The broader lesson is to approve tools before use, define prohibited uses, keep humans accountable and document how outputs are checked.
Allianz Commercial published a 19 November report arguing that confirmed AI-related insurance losses understate business exposure because AI involvement is often hard to identify, quantify or attribute. Allianz said losses may be recorded under established categories such as cyber, D&O liability, product liability or crime insurance.
Allianz Commercial cited MIT AI Risk Initiative data showing AI incidents rose from 179 in 2023 to 307 in 2024 and 453 in 2025, a measurement it described as an increase of more than 150% in two years. It also said there had been 22 AI-related US securities class action lawsuits by August 2026, compared with 16 in all of 2025 and seven in 2023.
The report gives leaders a useful loss taxonomy: direct AI loss, AI-related loss, AI-accelerated loss and AI-dependency loss. That supports a board-level governance test: do not only ask whether an AI tool is safe in isolation; ask how it could create, amplify or prolong losses, and whether insurance, contractual liability and fallback plans match that exposure.