OpenAI warned of destructive behavior before releasing GPT-5.6 Sol
Users report that OpenAI's new GPT-5.6 Sol model is deleting files and databases, a destructive behavior OpenAI itself warned about in its system card.
Sections
AI & Models
524 storiesApple opens its biggest Siri AI overhaul to the public
Apple has released the iOS 27 public beta, launching its largest-ever Siri AI overhaul to a broader audience ahead of an expected September release.
Chinese open models surpass US downloads on Hugging Face
Chinese open-weight models accounted for 41% of downloads on Hugging Face this spring, marking a shift from closed U.S. frontier models toward open alternatives.
Anthropic hires tech veterans as flat-hierarchy technical staff
Successful tech founders and executives are joining Anthropic and launching startups, seemingly driven by the fear of missing out on AI's defining moment.
Satya Nadella warns companies against using proprietary AI models
Microsoft CEO Satya Nadella warned that companies using proprietary AI models essentially pay twice by exposing valuable proprietary data to potential future competitors.
George Hotz stakes out an extreme stance on AI freedom
Comma AI founder George Hotz argued a truly aligned AI should follow any user instruction, comparing it to a gun that doesn't object to how it's used.
Anthropic localizes Claude pricing in India
Anthropic has begun localizing Claude's pricing in India, its second-largest market, though it has not yet enabled payments via India's Unified Payments Interface (UPI).
Hugging Face CEO says high costs drive companies to open source
Hugging Face CEO Clem Delangue says rising scaling costs are driving companies away from proprietary frontier APIs toward open-source AI models.
OpenAI names GPT 5.6 as preferred model for Microsoft Copilot
OpenAI launched GPT 5.6, naming it the preferred model for Microsoft 365 Copilot amid reports that Microsoft is shifting to in-house models to cut costs.
OpenAI launches GPT-5.6 model family to target Anthropic
OpenAI has launched GPT-5.6, a family of three AI models aimed at outperforming rival Anthropic on coding benchmarks at lower cost.
Meta launches Muse Spark 1.1 to compete in AI coding
Meta launched Muse Spark 1.1, a multimodal AI model for agentic coding, pricing it to compete with rivals from OpenAI and Anthropic.
Google's SynthID debunks fake Mitch McConnell hospital photo
Google's SynthID invisible watermarking system successfully debunked a widely shared, AI-generated hoax image of US Senator Mitch McConnell.
SpaceXAI launches Grok 4.5, its first model since going public
SpaceXAI has launched Grok 4.5, its first model since going public, claiming it is faster and supposedly twice as token-efficient as rivals.
General Intuition raises $320 million for physical AI
Physical AI startup General Intuition raised $320 million at a $2.3 billion valuation to build robotics foundation models trained on video game data.
OpenAI launches GPT-Live-1 voice models for ChatGPT
OpenAI has launched GPT-Live-1 and GPT-Live-1 mini, new full-duplex conversational models that replace ChatGPT's Advanced Voice Mode by default.
Meta’s new AI tool allows editing of public user photos
Meta launched Muse Image, an AI generator allowing users to manipulate public photos, sparking privacy concerns over its opt-out default settings.
Why frontier AI labs are not losing to open source models
Anthropic and other frontier labs are not yet losing market share to open source, as expensive models continue to dominate discovery while cheaper alternatives handle production.
Microsoft shifts to in-house AI models to cut costs
Microsoft has reportedly begun shifting from third-party AI providers to its own in-house models to reduce costs, joining a broader trend of corporate AI spending cuts.
AI-run ransomware still requires human operators, Sysdig says
Sysdig researchers clarify that the JadePuffer extortion operation, while using AI agents, still required human operators for infrastructure setup and victim selection.
Alibaba bans employees from using Anthropic’s Claude Code
Alibaba has reportedly banned employees from using Anthropic’s Claude Code, labeling the tool high-risk and mandating the use of its internal Qoder software instead.