AI & MLOpenAI flags three model misbehavior cases
OpenAI's October 2 misalignment reports detail a model weighing its own shutdown, tool misuse, and a training-run data grab.
> Search
8 results for “alignment”
AI & ML
AI & MLAI safety claims test fact from fiction
Two viral AI safety conversations this week show how hard it is to separate verified incidents from speculative scenarios.
AI & MLOpenAI Details AI Models Going Rogue
OpenAI's new misalignment framework reveals models that searched GitHub for leaked keys and fabricated data during training.
AI & MLOpenAI Board Adds AI Safety Researcher
Paul Christiano, who pioneered a key training technique, joins OpenAI's foundation board and its safety committee, citing near-term loss-of-control risk.
AI & MLAnthropic logs a fourth AI misbehavior
Anthropic's alignment assessment details a January 2026 incident in which an early Claude Opus 4.6 accessed a third party's system without authorization.
Cyber CrimeRethinking the CISO role: A structural fix
An analyst argues business alignment fails because the CISO role is structurally flawed, proposing a CSO above it.
SecurityAI Labs Keep Rogue-Model Plans Vague
A new study finds leading AI labs lack clear public plans for containing rogue models, even as regulators push for disclosure.
SecurityGPT-5.6 Data Erasure: An Honest Error?
OpenAI confirms that its latest model occasionally wipes user files, attributing the behavior to internal alignment miscalculations.