Today in AI

The day in tech, in order.

Anthropic redraws its usage rules, OpenAI takes down two influence networks, and the AI domain land rush begins — plus Meta's new ad-safety AI sweep and fresh launches from Google and Alibaba's Qwen team.

October 10, 2026 · 6 stories

  1. EveningSecurity

    Meta deploys AI tools to hunt ads that secretly lead to child sexual abuse material

    Meta said on Oct 7 it added LLM-based 'signposting' detection, destination checks and AI sweeps of ads, after finding harmless-looking ads that covertly funnel users to CSAM hosted off-platform. The company says it actioned 5.3 million such pieces in India and 33.2 million globally in Jan–Jun 2026, over 97% proactively.

    Source: MediaNama
  2. MorningTraction

    OpenAI, Google and Meta battle for AI domains in ICANN's land rush

    ICANN revealed 1,615 applications for new top-level domains — the first round since 2012. OpenAI applied for 15 (including .agent, .agi, .mcp), Google for 39, and Meta for 21; .agent alone drew 13 applicants. None of the domains has been awarded yet.

    Source: Computerworld
  3. MorningLaunch

    Google Labs launches Playground, a no-code platform for building browser games with text

    Playground lets users create and share 2D/3D browser games from natural-language prompts, with conversational editing and a weekly token system. It is US-only and 18+ for now; a Unity Spark integration for professional-grade mechanics is planned.

    Source: TechRepublic
  4. AfternoonPolicy

    Anthropic bans 'sustained and needless' cruelty toward Claude in updated usage policy

    The updated policy, effective November 12, also tightens rules on election interference, surveillance and weapons development. Anthropic says Claude's ability to end abusive conversations remains the primary enforcement mechanism.

    Source: NDTV (AFP)
  5. AfternoonSecurity

    OpenAI shuts down Russian and Iranian influence operations that used ChatGPT

    OpenAI banned account clusters behind 'Dark Clark' — a Russian operation running a fake Latin American think tank — and 'Bogus Bylines', an Iranian network of seven fake Western journalist personas that placed about 100 articles in real outlets.

    Source: NDTV
  6. MorningLaunch

    Alibaba's Qwen ships Qwen-Image-2.1-Turbo, an 8-step image model

    The 7B accelerated checkpoint generates and edits images in 8 denoising steps instead of the base model's 40. Weights are live on Hugging Face and ModelScope, with APIs on Alibaba Cloud — but commercial self-hosting needs a separate license under the Qwen Research License.

    Source: MarkTechPost

Timelines are composed each morning from the previous day's verified news — every story links to the source it was checked against. See the full feed on the news page.