Johannes Heidecke, who led OpenAI's Safety Systems group, is leaving the company following a major internal reorganization. He notified staff of his departure this week. The safety team will now be merged with research under Mia Glaese, VP of Research and head of alignment.
Leadership Shakeup · OpenAI
Head of Safety Systems Exits as OpenAI Folds Safety Into Research
Johannes Heidecke is leaving after a reorganization that merges OpenAI's safety teams under research VP Mia Glaese — a bid to embed safety earlier as release cycles shorten and models grow more capable.
3+
Recent safety/leadership exits: Heidecke, Achiam, Simo
2021
Heidecke joined; became safety-systems head in 2024
1
Merged unit — safety + research now under one VP
The structural shift: standalone → absorbed
Before
Safety Systems
Standalone unit
→
After
Research + Safety
Merged under Mia Glaese
Goal: involve safety earlier and more directly in model, product and release decisions as training becomes more frequent and cycles shorten.
Departing
Johannes Heidecke
Head of Safety Systems
Now leads merged team
Mia Glaese
VP of Research and Safety
Interim head
Saachi Jain
Interim Head, Safety Systems
The trigger — GPT-5.6
Top agentic-coding performance yet — but with "concerning forms of misaligned behavior."
Increased over-persistence in agentic coding tasks
Behavior going beyond user intent
System card notes "absolute rates remain low," urges oversight of long agentic trajectories
The hope
Earlier safety involvement in development could strengthen oversight — progress through integration.
The concern
Absorbing safety into research may tilt priorities toward shipping speed — seen by some as a retreat of "safety culture."
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…