The linked Dark Reading article is a much better read than the Axios offering (and isn’t behind a subscription wall):

https://www.darkreading.com/threat-intelligence/2026-agentic-ai-attack-surface-poster-child

The one to watch: Anthropic is privately warning top government officials that its not-yet-released model — currently branded “Mythos” — makes large-scale cyberattacks much more likely in 2026.

those humble brags disguised as warnings of the potential of future llm products are getting really old. in some ways for specific use cases llms arguably actually improved by a lot over the last three years. but in large its still garbage.