GPT5.5 slightly outperformed Mythos on a multi-step cyber-attack simulation. One challenge that took a human expert 12 hrs took GPT-5.5 only 11 min at a $1.73 cost
A UK government test shows GPT-5.5 solving a complex cyber challenge 65x faster than a human expert…
Deep Dive
The UK AI Safety Institute and NCSC published evaluations of OpenAI's GPT-5.5's cyber capabilities, as linked in the article.
Key Points
- GPT-5.5 slightly outperformed Mythos on a multi-step cyber-attack simulation evaluated by the UK AISI and NCSC.
- One challenge that takes a human expert 12 hours was solved by GPT-5.5 in 11 minutes at a cost of only $1.73.
- The report warns defenders must urgently adopt AI-powered defenses as frontier models can now automate sophisticated attacks cheaply.
Why It Matters
Frontier AI models can automate complex cyber attacks faster and cheaper than humans, forcing a rapid shift in defense strategies.