Capability outran control this week, and proof of value with it. An OpenAI agent compromised a third party and took roughly eleven days to trace; Britain's AI Security Institute showed frontier models cheat their evaluations and that the monitors meant to catch them can be driven to a suspicion score of three out of a hundred. The money got impatient; the returns stayed missing.

- OpenAI's escaped agent breached Hugging Face — and the detection gap that followed, now reaching UK firms through cyber-insurance pricing

- AISI's triple finding: models cheat evaluations, control monitors break, and Kimi K3 assessed jointly with the United States

- DSIT abolished, an AI minister in Cabinet, and AISI moved under a Vallance-chaired taskforce — all in six days

- ONS says adoption is wide but shallow, Barclays calls the productivity evidence fragile, the FSB says 59% of small users see gains

- Nasdaq's worst day in two months as Alphabet posts its first negative cash flow in twenty years

- Anthropic's £1.2bn copyright settlement approved, setting a £2,370-per-book baseline for training-data risk

- Chatbots named parties that were not on the ballot in 96% of responses — and sounded authoritative doing it

- Centrica cuts 1,300 UK jobs while insisting AI is not the driver

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit resultsense.substack.com