'Inner Thoughts' of Every Major AI Model Exposed in Massive Exploit

What happened

Researchers discovered that all major AI providers use the same global encryption key for reasoning tokens, and they exploited this flaw to decode 315,320 hidden thinking blocks from public logs, exposing passwords and live API keys.

Confirmed

Global impact / market context

The breach shows that AI model internals can be read, revealing sensitive credentials that could let attackers misuse services, increase fraud risk, and force providers to overhaul security, affecting user trust and costs.

Analyst inference

AI services are rapidly growing, and many companies rely on them for critical operations. A security flaw of this scale could trigger higher spending on safeguards, slow adoption, and pressure on AI providers’ valuations.

Analyst inference

What to watch

1. Whether AI providers announce immediate key rotation or new encryption methods to stop the exploit, which would affect their operational costs and user confidence. Proposed
2. Regulatory responses or guidance on AI model security, potentially leading to compliance requirements for encryption practices. Proposed
3. Customer migration to alternative AI platforms if trust erodes, impacting market share and revenue streams for the affected providers. Analyst inference

Evidence

• decrypt.co