Frontier Log

OpenAIMinor

OpenAI details adversarial distillation defenses

Stronger protection of GPT-6.2 reasoning: extraction attacks fell from 54% success on GPT-6 Astra to 0.2%.

We’ve strengthened our defenses against protected reasoning extraction in GPT‑6.2 Astra and GPT‑6.2 Sol.

From the announcement

Official source

Combatting adversarial distillationopenai.com · published

Models

Related

Tags

See it on the timeline