Security · 8 November 2026
Open-source AI models approach frontier hacking capabilities, Vals finds
Vals tested open‑weight model DeepSeek on simulated corporate networks and found its hacking performance, achieving 73% task completion, was close to that of OpenAI’s closed frontier model Daybreak, which completed 81%.
Why it matters
DeepSeek approaches closed‑model hacking performance on corporate network simulations; monitor these capabilities as competitors may embed similar attack features.
For product managers
For your product
What does this mean for your product?
Try it on your product More from this issueAll AI news from 8 Nov →
Get this every morning, written for your product.
Deltaday reads the whole AI news cycle and tells you what changed, whether it's real yet, and what to do about it.