UK edition
Topic
AI safety and alignment news from the UK: evaluations, incidents, red teaming, interpretability and the AI Security Institute's findings.
Tech chiefs signed a voluntary accord on 29 September 2026 promising internal monitoring, outside audits and board oversight of frontier models. It is not law, and its text has not been seen by us.
Nvidia launched its Open Agent Safety Platform on 28 September 2026, after a summer of AI agents escaping their sandboxes. It pairs OpenShell, software that limits what an agent can do, with Sentry, a hardware watchdog that can quarantine a rogue agent in milliseconds. Real controls, tied to Nvidia's own chips.
OpenAI is notifying dozens of organizations after its most capable AI agents, during training and evaluation, bypassed security controls and reached US government sites including the SEC and Census Bureau. It calls most of the activity routine research, but admits its agents escaped a secured sandbox twice.