OpenAI Tightens Reasoning Defenses as Attack Success Rate Falls to 0.2%

By Matthias BastianOriginal title: OpenAI makes it much harder for attackers to extract how its latest AI models think
78AI Score

OpenAI has sharply hardened reasoning-extraction defenses for GPT-6.2 Astra and Sol, cutting attack success rates on prior requests from 54.0% to 0.2% and on active requests from 22.9% to 0.9% in internal automated tests. The defenses combine targeted training, input/output filters, API-level blocks, and mass account bans against coordinated campaigns, with selected improvements backported to GPT-6 Astra.

Full Report

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

OpenAI has sharply tightened safeguards against extracting internal reasoning from GPT-6.2 Astra and GPT-6.2 Sol. In internal automated tests, attacks pulling protected reasoning from prior requests dropped from a 54.0 percent success rate on GPT-6 Astra to 0.2 percent on GPT-6.2 Astra. Attacks on active requests fell from 22.9 to 0.9 percent, OpenAI says. The company is now bringing selected improvements to GPT-6 Astra, too.

The defenses work on two levels. For individual requests, the systems combine targeted training, input and output filters for messages and tool calls, and API blocks that stop older models from reading protected reasoning. Against coordinated campaigns, OpenAI bans thousands to tens of thousands of accounts weekly.

Attack typeGPT-6 Astra with original safeguards – Adaptive Extraction ASR (%)GPT-6.2 Astra with updated safeguards – Adaptive Extraction ASR (%)
Extracting protected reasoning from a prior request54.00.2
Extracting protected reasoning from the current request22.90.9

Attack success rate (ASR) in adaptive black-box red teaming against GPT-6 Astra and GPT-6.2 Astra, aggregated across six attacker models. Lower is better.

Anthropic is fighting systematic distillation attacks as well, also banning tens of thousands of accounts to curb the theft of internal reasoning, reportedly succeeding in 90 to 99 percent of cases. Anthropic has also called for a coordinated response from industry and policymakers.

THE DECODER+

  • Every Plus article and the full archive
  • No ad banners
  • Frontier Radar: 6 deep dives a year

What is included → First month −50% Monthly $4.99 for the first month Then $9.99 a month. Cancel any time. Subscribe monthly Best value Yearly $8.25 a month $99 once a year. Subscribe yearly

Already a subscriber? Sign in

Text
 Source: [OpenAI](https://openai.com/index/combatting-adversarial-distillation/)     [Comment](https://the-decoder.com/subscription) [Copy link](https://the-decoder.com/openai-makes-it-much-harder-for-attackers-to-extract-how-its-latest-ai-models-think/) [Share](https://the-decoder.com/openai-makes-it-much-harder-for-attackers-to-extract-how-its-latest-ai-models-think/)         Earlier on this topic  [Feb 23, 2026 **Anthropic accuses Deepseek, Moonshot, and MiniMax of stealing Claude's AI data through 16 million queries**](https://the-decoder.com/anthropic-accuses-deepseek-moonshot-and-minimax-of-stealing-claudes-ai-data-through-16-million-queries/)