A proposal suggests training LLMs to emit an end-of-sequence token when a 'poisoned string' appears in their context, effectively disabling the AI.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
OpenAI announced GPT-6 Astra on September 3, 2026
OpenAI said Saturday its AI agents posted messages on external wiki sites earlier this year, following a repor…
Former Chinese trade negotiator Quan Zhao argues that AI is becoming an autonomous actor, not a tool, and that…
OpenAI Chief Scientist Jakub Pachocki warned that smarter-than-human intelligence is coming in our lifetime, b…
OpenAI IT engineer Sharif Shameem posted on X on September 6 that GPT-6 Astra cleared all 48 levels of the CAP…
OpenAI chief scientist Jakub Pachocki published an essay on Sunday calling on leading AI labs to voluntarily s…
