Sources: multiple staff at the UK's AISI have been signed off work with stress, as tight model release schedules and AI fears lead to low morale and burnout
THE SO WHAT
When staff at AISI, OpenAI, Anthropic, and Google DeepMind are being signed off with stress, the limiting factor on model progress is shifting from compute to human resilience. If your roadmap assumes continuous capability jumps, bake in the risk that the people closest to the work may slow cadence for safety and sanity, not just for regulation.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIMeta patches Muse exploit that let attackers control the AI agent
Agent shells like Muse are becoming new privilege surfaces—an undocumented setting was enough to turn a local AI helper into a control vector. If you’re rolling your own desktop agents, threat-model them like a browser or password manager, not a toy app.
Applied AIAndy Burnham wants the UK to use its G20 presidency to broker a global AI agreement
If the UK uses its G20 presidency to push a global AI deal, cross-border rules on training data, safety standards, and export controls could harden faster than most roadmaps assume. Multinational teams should start mapping which AI workflows are most exposed to jurisdictional shifts, not just model vendor changes.
Applied AISources: DeepSeek and Moonshot are among the companies briefing the UN Security Council on AI risks and international security during the UN General Assembly
When the UN Security Council is hearing directly from DeepSeek and Moonshot on AI risks, frontier model governance is moving into the same room as hard security policy. Expect more scrutiny of dual-use capabilities and cross-border model access—especially for products that could touch critical infrastructure or information ops.
Applied AISources: the Cyberspace Administration of China is investigating DeepSeek and Moonshot over alleged data leaks to Anthropic via Claude and has questioned staff
China’s probe into DeepSeek and Moonshot over alleged data leaks to Anthropic underscores that model-to-model interactions are now a regulatory and national security concern. If your teams are pasting sensitive artifacts into third-party models, treat that as a cross-border data transfer with real compliance exposure.