Reuters reported that Americans for Responsible Innovation urged the US administration to require safety reviews for frontier AI models before public release. The proposal would link government contract eligibility to CAISI-led review of cyberattack and weapons-development capabilities. 

The Guardian reported that Google’s threat intelligence group warned that AI-powered hacking has become an industrial-scale threat. The report says criminal and state-linked actors are using commercial models to refine attacks, scale operations and search for exploitable vulnerabilities. 

Regulation

  • Council of the European Union approved conclusions calling for a human-centred approach to AI in education. The conclusions focus on teacher skills, AI literacy, inclusion, fairness, wellbeing and safeguards against bias, misinformation and data-protection risks. 

  • UK National Screening Committee opened a consultation on evidence for automated AI grading in the diabetic eye screening programme. The evidence summary says some automated retinal image analysis systems may match first human graders, while performance varies by system, setting, ethnicity and age. 

  • National Cyber Security Centre published guidance on questions to ask before using AI models to find vulnerabilities. The guidance urges organisations to define the security objective, manage AI-found vulnerabilities, assess risks and verify results with both AI and humans. 

Cases

  • ABC7 San Francisco reported same-date live updates from Musk v Altman, with Microsoft CEO Satya Nadella expected to testify as week three of the federal trial began in Oakland. The proceeding remains relevant to AI governance because it concerns OpenAI’s nonprofit mission, control structure and alleged departure from founding commitments. 

Academia

  • Springer published Formation-based AI ethics: ethical formation, responsibility, and opportunity cost in AI ecosystems in AI and Ethics. The article argues that principle-based and tool-based approaches are insufficient unless AI governance also builds moral capacity across design, deployment and organisational practice. 

Events

  • IAPP will hold AI Governance Global Europe 2026 in Dublin from 1 to 4 June 2026. The conference focuses on AI governance challenges, emerging trends, best practice and upcoming legislation, with sessions on agentic AI, contracting and risk management. 

  • Oxford Faculty of Law will host Lecture: The EU AI Act, Ethics, and Healthcare on 26 May 2026. The event examines the EU AI Act through human agency, trust, risk, soft law, accountability and medical-device regulation. 

Sources: Reuters, The Guardian, Council of the European Union, UK National Screening Committee, National Cyber Security Centre, ABC7 San Francisco, Springer, IAPP, Oxford Faculty of Law

Tags
caisi
hacking
teachers
screening
vulnerabilities
OpenAI
Healthcare