AI Tools Cut Both Ways in Security; Cowork Leaves the Desktop

Popular AI tools as attack infrastructure

Ars Technica reports that hackers can use nine of the most popular AI tools to assemble massive botnets. The claim is about what the tools make possible, which is narrower than documented large-scale abuse. Still, for engineers who build on or operate these services, it suggests AI platforms may deserve a place in the threat model alongside other services an outsider can drive at scale.

Defensive work is moving too, though not in response to that report. Cotool has published research on applying AI to blue team operations, centered on defensive strategies against cyber threats. The company argues that bringing AI into security protocols can strengthen defenses and speed up incident response.

Meta and ByteDance release new models

Meta's Superintelligence Labs has launched Muse Image, its first AI image model, across the Meta AI app, Instagram and WhatsApp. Shipping directly into those apps gives the model immediate reach for image generation and editing. ByteDance, meanwhile, debuted Seedream 5.0 Pro, which it pitches on advanced reasoning. The two launches differ in emphasis: Meta is leading with distribution inside products people already use, while ByteDance is leading with the model's reasoning ability.

Claude Cowork moves beyond the desktop

Anthropic is expanding Claude Cowork from desktop to web and mobile. The new versions are in beta and limited to Max subscribers for now, so most users will not see them yet. For people who already rely on Cowork at a desk, the change means the same assistant can follow them onto a phone or into a browser session.