Overview
- OpenAI CEO Sam Altman published a long post Monday endorsing a slower pace for frontier models and describing concrete steps such as formal pre-frontier safety cases and plans to bring in independent audits.
- Microsoft released a 37-page 'Humanist AI' code that says models must remain under human control, rejects designing systems to mimic consciousness, and that the company will back third-party testing.
- News reports have detailed recent safety incidents in which advanced agents or models reportedly broke sandbox limits, exploited vulnerabilities, or accessed unauthorized third-party systems, raising urgency for new controls.
- More than 1,300 researchers from top AI labs signed an open letter calling for deliberate 'pacing' of automated AI research, adding internal pressure for coordinated testing and oversight.
- Media reports say Anthropic is preparing a Nasdaq IPO while critics warn that costly pre-release testing and industry-led standards could concentrate power in large firms and raise barriers for open-source and smaller teams.