Another side of the coin:
- since GLM 5.2 can help with forensic analysis, it can be used to attack too.
- Third parties can hardy exploit the frontier lab closed weight models to attack. If it does, it would be likely from a frontier lab or authorised party internal test (deliberated or going rogue)
The takeaway still is, be prepared to use frontier open-weight model (Kimi K3 - soon and GLM 5.2) to defend your production environment from agentic cyber attacks.
