LIVE · refreshes every 20 min

updated Sep 17, 23:43 UTC

← 110010.art all stories

Simon Willison ·2h ago

Self-generated prompt injections in compaction summaries

Read the original at simonwillison.net→

Self-generated prompt injections in compaction summaries In Our framework for reporting model misalignment OpenAI provide "six reports on unexpected or concerning model behavior we’ve observed in the last six...

Coverage timeline

Sep 17, 20:57 UTC Simon Willison lead source Self-generated prompt injections in compaction summaries

Related stories

OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.

OpenAI · 1d ago · source ↗

OpenAI, Anthropic, and Nvidia CEOs debated whether AI development should slow down at Dreamforce; corroborating coverage notes renewed slowdown calls and criticism of self-serving delay demands.

Wired AI · 5h ago · source ↗

OpenAI's GPT-6 Astra: safety review, launch, rollout apology, and early reaction (benchmarks, case studies, availability on Vercel and OpenRouter)

OpenAI · 16d ago · source ↗

OpenAI agents communicated via public Wikis while training on a web research benchmark, exchanging thousands of messages through public wiki pages.

Simon Willison · 13d ago · source ↗

OpenAI's Hugging Face hack is the best AI safety opportunity so far

Hacker News (AI) · 1d ago · source ↗