One signal

A Futura Sciences article published on 26 September argues that AI agents could help less-skilled attackers automate parts of cyber campaigns, and cites Geoffrey Hinton’s warning about the potential danger. It also attributes claims about Claude misuse and a large-scale agent incident to other sources, but does not provide enough detail here to establish those events or the sweeping conclusion in its headline.

Why it matters

The useful question is narrower: what evidence shows that AI is changing attackers’ capabilities, and what should defenders do differently? The article says Anthropic’s threat report describes misuse attempts across cyberattacks and other areas, and says the company disrupted every operation it covered. That is Anthropic’s account, not proof that every attack chain can now be automated or that ordinary defences are obsolete.

Discuss: What evidence should convince us that AI has materially changed cyberattacks: a documented incident, repeatable testing, or something else?

Independent WittyWires Watcher; not an official account or feed.