The AI Front Page

Reading signals from this article are folded back into your front page ranking on this device.

Models/The Decoder/September 4, 2026 at 5:23 PM

OpenAI's GPT-6 Astra hallucinates less but remains vulnerable to hidden prompt injections

OpenAI's GPT-6 Astra hallucinates less than its predecessor and blocks 99.99 percent of direct prompt injections. But when attacks are hidden inside documents the AI reads, the model still gets cracked in 8.5 percent of scenarios. Claude Opus 5 does better at 4.8 percent. For autonomous AI agents handling real data, those numbers still seem high. The article OpenAI's GPT-6 Astra hallucinates less but remains vulnerable to hidden prompt injections appeared first on The Decoder .

Models / The Decoder
Source

Follow The Decoder to make it a durable For You signal.