cd /news/ai-safety/openai-makes-it-much-harder-for-atta… · home › topics › ai-safety › article
[ARTICLE · art-173480] src=the-decoder.com ↗ pub= topic=ai-safety verified=true sentiment=↑ positive

OpenAI makes it much harder for attackers to extract how its latest AI models think

OpenAI cut the success rate of attempts to extract protected internal reasoning from its GPT-6.2 Astra and GPT-6.2 Sol models from 54 percent to 0.2 percent in adaptive red-teaming tests, The Decoder reported. OpenAI also said it suspends thousands to tens of thousands of accounts each week over coordinated distillation campaigns.

recommendation

Red-team your models adaptively for reasoning-extraction attacks and monitor for coordinated distillation campaigns; OpenAI's adaptive testing cut extraction success from 54 percent to 0.2 percent.

more in playbook: Securing agents and AI tools
by read1 min views5 publishedNov 20, 2026

OpenAI has significantly strengthened safeguards against extracting internal reasoning from GPT-6.2 Astra and GPT-6.2 Sol. In adaptive red-teaming tests, the success rate for extracting protected reasoning fell from 54 percent to 0.2 percent. The company is also suspending thousands to tens of thousands of accounts each week over coordinated distillation campaigns.

The article OpenAI makes it much harder for attackers to extract how its latest AI models think appeared first on The Decoder.

── more in #ai-safety 4 stories · sorted by recency
── more on @openai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/openai-makes-it-much…] indexed:0 read:1min 2026-11-20 · —