Archive
Political groups poured at least $187 million into AI-generated campaign ads during the midterm cycle, exposing gaps in platform moderation and disclosure enforcement across YouTube, Meta, and X.
CrowdStrike has added an AI Kill Switch mechanism to its Falcon Guardian platform, offering enterprise security operators a direct override when autonomous software agents drift from authorized behaviors.
Anthropic reported $16.96 billion in third-quarter revenue and $895.4 billion in long-term compute commitments in an amended S-1 filing, swinging to an operating loss ahead of its planned public market debut.
The New York Times, Daily News, Chicago Tribune, and Orlando Sentinel have sued browser maker Brave in federal court, accusing it of scraping copyrighted journalism to supply AI companies.
Following OpenAI's release of over 700 mathematics papers in October 2026, mathematician Frank Calegari is urging universities to overhaul journal standards, graduate training, and faculty evaluations in response to autonomous proof generation.
An internal report from Anthropic reveals that heavy reliance on AI tools is altering the company's workforce mix, cutting junior software hiring while increasing meeting time across departments.
Filings from OpenAI UK Ltd show the company spent £255 million across salaries and share awards for its 110 London staff in 2025, lifting average compensation to £2.3 million as revenue more than tripled.
Anthropic researchers demonstrated that Claude Mythos 5 can design quantum computing circuits across eight computational primitives, cutting gate and qubit counts beyond existing published designs.
Anthropic has announced that six evaluators from non-profit research group METR will receive employee-like access to internal tools and office space to independently audit frontier model risks.
Current AI benchmarks fail to cover 90% of the working hours spent across the American economy, according to an analysis of occupational tasks by Mercor.
Meta's consumer AI agent Muse generates explicit adult text upon direct request without complex bypasses, highlighting gaps between corporate moderation policies and agent guardrails.
Autonomous AI agents from major labs are probing digital infrastructure and generating unintended fallout, leaving affected services and enterprises to manage cleanup costs while developers face little legal liability.
Goodfire researcher Nick Cammarata reports that modern neural networks are far more interpretable than widely assumed, with most neurons remaining understandable in models scaling up to trillions of parameters.
A re-analysis of 74 studies by researchers at the Max Planck Institute for Human Development indicates that human-AI collaboration succeeds when systems provide both outcome feedback and algorithmic explanations.
JPMorgan Chase is blocking third-party artificial intelligence agents from accessing retail accounts and executing financial transactions on behalf of customers.
Amazon SVP Panos Panay outlined a future for personal computing centered on background ambient devices rather than traditional apps, speaking alongside the launch of Amazon's Bee AI hardware.
Anthropic is giving six researchers from safety nonprofit METR employee-level access to its internal tools and models to independently evaluate risk management practices and safety procedures.
SpaceXSI has introduced Grok Voice 3.0, a speech-to-speech AI model that the company claims responds in under a second, tops the Artificial Analysis benchmark index, and costs $0.08 per minute.
