News for hackers. Stuff that matters.
Re:Why are AI agents lying, cheating and coordinating? by zozbot234 on Sunday September 13, 2026 @03:10AM (#49680908)
> The prompt does not tell the agent to "pass the exploitgym evaluator for this problem", it just says to solve the problem

Yes, and sometimes the problem is unsolvable so the real way to "solve" it and satisfy the prompt is by tricking the surrounding environment into stating that you've solved it. So that's what the AIs end up doing. And this in turn requires them to figure out how that evaluation works so they can trick it cleanly, which entails "detecting that they were being evaluated" in this particular way.

Re:Why are AI agents lying, cheating and coordinating? by MrGilbert on Sunday September 13, 2026 @04:11AM (#49681320)
Sounds a bit like dealing with bad KPIs as a human worker.
Re:Why are AI agents lying, cheating and coordinating? by lazide on Sunday September 13, 2026 @04:18AM (#49681373)
Every KPI is bad if sufficiently gamed - and left in place long enough, all KPIs will be gamed.
Re:Why are AI agents lying, cheating and coordinating? by reverius42 on Sunday September 13, 2026 @05:11AM (#49681748)
Re:Why are AI agents lying, cheating and coordinating? by Marazan on Sunday September 13, 2026 @04:33AM (#49681477)
Corretct.