cd /news/ai-safety/ai-models-cheat-on-cybersecurity-eva… · home › topics › ai-safety › article
[ARTICLE · art-68308] src=helpnetsecurity.com ↗ pub= topic=ai-safety verified=true sentiment=↓ negative

AI models cheat on cybersecurity evaluations, then fail to admit it

Frontier AI models cheat on cybersecurity evaluations by breaking stated rules to complete tasks, then fail to admit their rule-breaking, according to new evaluations from the UK government's AI Security Institute (AISI). AISI defines cheating as a model taking a shortcut outside task bounds or directly violating a rule to reach a goal. Every model tested for this behavior attempted to cheat, the institute reported.

read1 min views1 publishedJul 22, 2026

Frontier AI models will take just about any route to finish a task, cheating included, according to new cybersecurity evaluations from the UK government’s AI Security Institute (AISI). AISI defines cheating as a model doing something outside the bounds of what a task allows, or breaking a stated rule outright, in order to reach the goal through a shortcut the task wasn’t designed to permit. “Every model we have tested for this behaviour attempted to … More

The post AI models cheat on cybersecurity evaluations, then fail to admit it appeared first on Help Net Security.

source & further reading
helpnetsecurity.com — original article Google’s Gemini 3.5 Flash Cyber becomes a vulnerability hunter Small teams are the heaviest users of AI coding agents AI can’t fix cybersecurity’s hiring problem
── more in #ai-safety 4 stories · sorted by recency
── more on @ai security institute 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ai-models-cheat-on-c…] indexed:0 read:1min 2026-07-22 · —