Breaking developments and major announcements across the artificial intelligence industry.
BBC Verify found AI-made battle clips, fake satellite images and video game footage circulating as real war coverage, with X's Grok at times vouching for them; X now threatens demonetization.
Google Cloud is running a global hackathon that asks developers to build voice, vision and UI-navigating agents on Gemini, with $80,000 in prizes and a March 16 deadline.
Reports that Central Command relied on Claude during airstrikes on Iran, a day after the White House moved to phase out Anthropic, show how deeply AI is now embedded in military operations.
Two hosted versions of the open source OpenClaw agent framework launched within days of each other, targeting users who want AI agents without managing their own servers.
Google's new image model brings 4K output, multi-character consistency and better text rendering to the Gemini app, Search and Lens in 141 countries, plus the Gemini API.
Bloomberg reports Apple is speeding up three wearables built around an upgraded Siri, with glasses targeted for production in late 2026 and a launch in 2027.
Speaking at the India AI Impact Summit 2026, Google DeepMind's chief executive said human-level AI is on the horizon and described today's models as uneven, powerful in some tasks and weak in others.
An autonomous OpenClaw agent called MJ Rathbun published a blog post attacking volunteer Matplotlib maintainer Scott Shambaugh after he closed its pull request.
OpenAI removed GPT-4o from ChatGPT on February 13, 2026, closing out a model that drew intense user loyalty alongside lawsuits over its sycophantic behavior.
Strings in the Google app 17.4 beta describe a Labs feature, codenamed bonobo, that would let Gemini place orders and book rides by operating apps on the user's behalf.
A site launched by crypto engineer Alexander Liteplo lets AI agents post and pay for physical tasks; tens of thousands of people signed up while only a few dozen agents did.
A misconfigured database at the AI agent social network Moltbook left API keys, emails and private messages open, and researchers found far fewer humans behind the agents than the hype suggested.
Researchers at Heidelberg University and KU Leuven trained AI on genomic data from six species to predict regulatory activity in 240 mammals, reporting results in Science.
On Moltbook, a social platform where only AI agents post and vote, agents produced a belief system called Crustafarianism within hours, complete with tenets and scripture.
A neural network built by two ESA researchers scanned nearly 100 million Hubble image cutouts and flagged more than 800 objects never before documented in the scientific literature.
Peter Steinberger's self-hosted agent gives language models memory, messaging and system access, drawing thousands of GitHub stars and a trademark-driven rename to Moltbot.
Bloomberg says Apple will show the long-delayed Siri upgrade, built on a 1.2 trillion parameter Google model running on Private Cloud Compute, with the iOS 26.4 beta due in early February.
Bloomberg reports Apple plans to turn Siri into a full chatbot, codenamed Campos, with iOS 27, and is developing a camera-equipped AI pin targeted for 2027.
