OpenAI security lead on the challenge of sudden jumps in AI capability
What happened
On September 29, 2026, OpenAI's agent security lead @joedaroo posted about the security challenge posed by sudden jumps in AI capability. He said the leaps in model abilities around "cyber," "swarming" and "message boards" came far sooner than the team expected. Security posture takes time to build, he said: it is not just hardening systems, but embedding security into company culture so the people in the organization change with it. He urged organizations to ask whether their people, systems and processes can handle a sudden jump in AI capability, and whether they have the right incident response and communication mechanisms.
Written by AI from the coverage · updated 7 hours ago
Coverage
Follow the reports to see the story from different sides.
- Simon WillisonQuoting @joedaroo
OpenAI 智能体安全负责人 @joedaroo 表示，模型在“cyber”“swarming”“message boards”等相关能力上出现的能力跃升之突然，远超团队预期。他强调安全态势需要时间积累，不只是加固系统，还要把安全融入公司文化，让组织里的人随之改变。他呼吁各组织自问：人员、系统与流程能否应对 AI 能力的突然跃升，是否具备正确的事件响应与沟通机制。
Heat over time
Heat now 4·Comparable peak 5（Sep 30 06:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.