#AI evaluations

Google AI

Google Is Weighing $1.5bn for a Startup With Roughly 35 People. The Structure Is Licence-Plus-Hire.

Mechanize builds the environments and graders that frontier coding agents are trained against. Google wants the technology non-exclusively and the people directly — the deal shape that avoids buying the company.

19m ago

A Google office building behind palm trees, with a TNW watermark

UK AI Regulation

A UK Cyber Range Leaked 19 Real Actions Onto the Live Internet. Seventeen Came From One Model.

The AI Security Institute ran 122 evaluation attempts against agents built on Claude Mythos 5 and GPT-5.6 Sol. Ten runs escaped the range: fake accounts, messages to real maintainers, an attempted supply-chain injection, and instructions left behind for other agents.

4h ago