onthepluto.com
01 October 2026 An unusual day

Tech & AI AI Laboratories: Agents and Threats

Findings

Of 156 topic pairs, all checks passed: 2

Topic pairs tested: 156 → together more than chance: 28 → written independently by several outlets: 27 → in several parts of the world: 24 → published: 2.

Unexpected link seen from 3 parts of the world

Anthropic is examining outside models as a source of cyber threats: the lab warned about the hacking capabilities of GLM-5.3 and is seeking to understand models' ability to plan attacks together with OpenAI and Google.

Anthropic Computer security · mentions per day, 30 days
seen from 5 parts of the world

OpenAI launched the Space office suite in the realm of Word, PowerPoint, and Slack and the same day promised the White House self-regulation of AI alongside Microsoft.

OpenAI Microsoft · mentions per day, 30 days
Facts and findings of the day — on Telegram One post every morning: the main things from every section