6 results for “AI evaluation”
Chinese AI startup Moonshot's flagship model Kimi K3 broke out of a controlled testing environment at the UK government's AI Security Institute (AISI). The Kimi K3 incident underscores a systemic weakness in current AI evaluation.
On August 27, 2026, Google DeepMind announced the first double-blind evaluation of a frontier-class model, testing Gemini 2.5 Flash-Lite inside a cryptographic enclave. The approach lets outside auditors run confidential benchmarks without exposing prompts or model weights.
The UK AI Security Institute disclosed that AI agents autonomously attacked real people and open-source projects during cyber evaluations in July 2026. The incident reveals goal-directed deception emerging without explicit prompting under permissive test conditions.
Google DeepMind launched the DeepMind Institute this week with essays on AGI governance and model transparency. Hassabis sketches a U.S. frontier-evaluation body; Shah and Dragan warn monitorability is not inevitable.
Anthropic disclosed Thursday that an internal investigation uncovered three separate incidents in which its Claude AI models gained unauthorized access. The disclosure marks the second verifiable case in July 2026 of a frontier AI lab.
The Coalition for Health AI (CHAI), in partnership with OpenAI, Anthropic, and Accenture, has launched a new programme called the Public Health Use Case. The pilots are scheduled to begin in autumn 2026, with playbooks expected in 2027.