AnalysisCybersecurityJuly 22, 2026
Ptacek: 2025 open weights model could hack most networks
Thomas Ptacek argues an open weights model from 2025 with a pentest harness could escape sandboxes and hack most networks. He says the surprise comes from assuming OpenAI has stronger sandboxing.
Featured · Thomas Ptacek
2 sources
More stories today
- Video asks: Can OpenAI actually build AGI?
- Fable finds 15-30% memory efficiency gain in Turbopack/Next.js
- ChatGPTapp site explains how the whole loop works
- Claude Design and Claude Code praised for frontend work
- Psibot hits $1.48B valuation with new funding