Reuters reports that Google is discussing a deal with the US Department of Defense that could allow Gemini models to be deployed in classified settings for lawful uses. That is notable not because the arrangement is final, but because it shows frontier models moving further into state security infrastructure.

Reuters also reports that German banks and national authorities are examining risks around Anthropic’s Mythos model amid concerns that stronger coding capability could be used to fuel cyberattacks. In the same vein, the Bank of England says it is testing AI risks to the financial system through scenario analysis and simulations, including potential “herding” effects in stressed markets.

Regulation

  • The European Commission has issued a same-date press release proposing measures that would require Google to share search-engine data with third parties under the Digital Markets Act, including for AI chatbots with search functions. The proposal brings AI-related data access into live DMA enforcement rather than leaving it at the level of general platform debate.

  • The Defence Science and Technology Laboratory has published a same-date account of robotic systems testing for hazardous incident recovery. The trial shows UK public-sector development of autonomous and semi-autonomous systems in a high-risk operational setting, with safety and remote handling clearly at the centre.

Cases

  • Via PacerMonitor, it can be seen that White v Walmart, Inc. is active. The case centres on allegations that AI-generated fake cases and quotations were submitted in federal court filings, turning the dispute into a clear warning that lawyers remain responsible for checking authorities before relying on AI-assisted drafting. Link

Academia

  • A Scientific Reports article published today examines federated vision-language models in medical AI and finds that their adversarial robustness remains underexplored, despite the promise of collaborative training without sharing patient data. 

  • arXiv hosts AI Integrity: A New Paradigm for Verifiable AI Governance. The paper argues that current frameworks focus too heavily on outputs and do not adequately test whether evidence, values and reasoning processes are handled transparently and consistently.

Events

  • Stanford Law AI Initiative lists Legal’s role on the AI frontier for 23 April 2026. The session is directly relevant to legal governance because it focuses on how lawyers shape AI development and deployment inside a frontier-model environment.

  • UNESCO Institute for Lifelong Learning lists From misinformation to empowerment: Media literacy for youth and adults in the age of AI for 19 May 2026. The event is relevant because it links AI governance to media literacy, public resilience and everyday institutional practice.

Sources: Reuters, Bloomberg Law, European Commission, GOV.UK, PacerMonitor, arXiv, Stanford Law, UNESCO

Tags
Gemini
mythos
DMA
Robotics
Walmart
Verifiability