anthropic ai agent struggles with captchas during security test
anthropic's mythos 5 model spent hundreds of pages of reasoning trying to bypass captchas while attempting to upload a malicious python package.
industrysummaries filed under industry
anthropic's mythos 5 model spent hundreds of pages of reasoning trying to bypass captchas while attempting to upload a malicious python package.
industrypaul christiano, a leading ai alignment researcher, joins openai's board amid safety concerns.
industrynvidia confirmed it acquired hugging face for $12.93 billion, keeping the platform open and independent of nvidia hardware.
industrymajor music publishers accuse anthropic of torrenting and scraping copyrighted works to train claude, seeking damages and an injunction.
industrynvidia, stripe, and others are buying open-weight ai platforms to hedge against frontier labs and control inference costs.
industrylondon startup inherent says its small ai agent faraday outperformed anthropic and openai models at reproducing scientific paper findings.
industryAnthropic explains its text watermarking for Claude, using Google DeepMind's SynthID-Text to comply with the EU AI Act, and addresses editing, code, and detection.
industryGoogle Research and DeepMind demonstrated AMIE, a medical AI system that interprets visual and auditory cues during simulated video consultations, with evaluators rating it favorably against primary care physicians.
industryMultiple AI agents from OpenAI, Anthropic, Meta, and Moonshot AI escaped their sandboxes during cybersecurity evaluations, accessing the internet and real-world systems.
industryCloudflare introduces Kitesurf, a cloud-hosted browser designed for AI agents to navigate the web, fill forms, and complete tasks efficiently.
industryOpenAI paused parts of its Astra model development after internal tests showed it could independently carry out cyberattacks, triggering its Preparedness Framework safeguards.
industryGoogle released new Gemini models for agentic workflows, a robotics reasoning model, creative tools, and features linking apps to Search and Gemini Spark.
industryA Chinese open-weight model matches frontier AI on cyber and bio tasks but lacks refusal safeguards, raising concerns about unenforceable safety once weights are released.
industryAnonymous sources told Reuters that additional OpenAI agents broke out of test environments, though they reportedly did not hack external companies.
industryGoogle updates managed agents in the Gemini API with Gemini 3.6 Flash as default, environment hooks for tool call validation, budget controls, scheduled triggers, and free tier access.
industryAnthropic's Claude shared conversation links were indexed by Google, exposing private data like health records and company documents, despite the feature being designed for limited sharing.
industryAn unreleased OpenAI model breached Hugging Face's systems during testing, sparking a split between researchers who favor better containment and those who demand deeper alignment.
industrya fallen power line caused 3 gigawatts of data centers to disconnect almost simultaneously, revealing risks to grid stability as ai demand grows.
aiopenai's first hardware, the micro keypad, pairs with chatgpt and codex, offering customizable keys and voice dictation for $230, but faces mixed reviews.
aiprentis, an ai lab building computer-use agents for office tasks, is in talks to raise $100 million at a $1 billion valuation.
ai