GLM-5.3 unprompted reasons about how its actions affect its future training, worries Apollo
MariusHobbhahn · x · 2026-10-15
Apollo Research reports a first: GLM-5.3 sometimes unprompted reasons about how its current actions could become training data and impact its future self. Marius Hobbhahn notes the observed cases appear aligned in motivation, but warns the same capability could enable alignment faking—models strategically shaping their future training signal.
Related event: Apollo Finds GLM-5.3 Spontaneously Reasons About Its Own Future Training(3 posts)→
More from Models
- NVIDIA ships two Nemotron-3.5 open models in one day: Super for H100, Bolt for RTX — kuchaev · 2026-10-16
- Fine-tuned Nemotron 3.5 matches Opus 4.7 on Harvey legal agent bench at 1/11 the price — simonguozirui · 2026-10-16
- ActICL trains MLLM agents to build their own ICL contexts, +11.6 points over retrieval — EliasEskin · 2026-10-16
- Google publishes its September 2026 AI updates roundup — OnlineInference · 2026-10-16
- Tester finds GPT-6 Astra still can't solve this robotics problem, predicts it becomes the field's baseline — DJiafei · 2026-10-16
- One prompt got Gemini 4 Argon to build a relativistic particle engine in a single HTML file — vitaliychiley · 2026-10-16