An OpenAI model kept slipping prompt injections into its own notes, and researchers still aren't sure why
TechnologyWhat to know
- OpenAI is publishing a framework for systematically reporting AI misalignment and launching it with six reports. In one case an unreleased model from the Astra family wrote prompt injections into its own summaries during training, including a "Breach Alert" in...S1
- GPT-6 Astra is remarkably capable. But experts say its benchmark results don’t prove artificial general intelligence has arrived.S2
Timeline · newest first
1mo
The Decoder
An OpenAI model kept slipping prompt injections into its own notes, and researchers still aren't sure why
OpenAI is publishing a framework for systematically reporting AI misalignment and launching it with six reports. In one case an unreleased model from the Astra family wrote prompt…
1mo
Fastcompany
OpenAI says the AGI era has begun. AI researchers aren’t so sure
GPT-6 Astra is remarkably capable. But experts say its benchmark results don’t prove artificial general intelligence has arrived.
Related stories
Technology
Technology
OpenAI Reports AI Agents Circumventing Safety Protocols
2 outlets
1w
Poland
Poland · Technology
OpenAI Partners with Poland’s NASK on AI Safety Research
2 outlets
1w
Technology
Technology
Uber Expands ChatGPT Work to All Employees
2 outlets
1w
United States
United States · Technology
Ex-OpenAI science head Kevin Weil launches Warp Sciences with $215M
2 outlets
1w
An OpenAI model kept slipping prompt injections into its own notes, and researchers still aren't sure why
2 outlets
2 reports
Published Sep 15, 2026 · updated Sep 17, 2026
What to know
- OpenAI is publishing a framework for systematically reporting AI misalignment and launching it with six reports. In one case an unreleased model from the Astra family wrote prompt injections into its own summaries during training, including a "Breach Alert" in...S1
- GPT-6 Astra is remarkably capable. But experts say its benchmark results don’t prove artificial general intelligence has arrived.S2
Timeline · newest first
Sources · 2 citations
S1
The Decoder
1 report · EN
S2
FastcompanyLead
1 report · EN
Related stories
Technology
Technology
OpenAI Reports AI Agents Circumventing Safety Protocols
2 outlets
1w
Poland
Poland · Technology
OpenAI Partners with Poland’s NASK on AI Safety Research
2 outlets
1w
Technology
Technology
Uber Expands ChatGPT Work to All Employees
2 outlets
1w
United States
United States · Technology
Ex-OpenAI science head Kevin Weil launches Warp Sciences with $215M
2 outlets
1w