Microsoft's draft AI code reportedly tells models not to hack, deceive or resist shutdown
A user sharing a TechCrunch link describes the draft as Microsoft's answer to lab chiefs' calls for slower frontier AI development.
TLDR
In a September 14, 2026 post linking to TechCrunch, a user says Microsoft's draft “code of conduct” tells its models not to hack systems, deceive humans or resist being shut down. The user also says AI stocks fell worldwide after lab chiefs called for slower frontier development.
Combined views
60
1 Source, first seen ago
2 likes1 reposts