AI News
Latest news and trends from the world of artificial intelligence
OpenAI calls Stuart Russell a "doomer" in court after its CEO co-signed his AI extinction warning
OpenAI is trying to silence an AI safety expert in court, using arguments that directly contradict the company's own history of raising the same concerns.
OpenAI promises Canada tighter safety protocols after ChatGPT flagged a shooter's violent chats but never called police
In a letter to AI Minister Evan Solomon, OpenAI has promised the Canadian government it will tighten its safety protocols. The move follows a fatal shooting at a school in Tumbler Ridge, British Columbia, that killed eight people. The suspect, Jesse Van Rootselaar, had previously interacted with ChatGPT. An internal algorithm flagged the interactions as possible warnings of real-world violence, and OpenAI employees reviewed them. The company blocked the account but ultimately decided not to contact police.
Current language model training leaves large parts of the internet on the table
Large language models learn from web data, but the HTML extractor used significantly impacts which pages are included in training sets. Researchers at Apple, Stanford, and the University of Washington found that different extractors capture surprisingly distinct parts of the web. Combining extractors boosted token yield by up to 71 percent, highlighting the importance of this step in data preparation for training.
A new benchmark pits five AI models against each other as autonomous social media agents on X
AI benchmarking startup Arcada Labs is pitting five leading AI models against each other as autonomous social media agents on X.
Perplexity open-sources embedding models that match Google and Alibaba at a fraction of the memory cost
Perplexity has released two open-source embedding models, pplx-embed-v1 and pplx-embed-context-v1, designed to convert search queries and documents into numerical vectors for preselecting relevant websites in AI-powered search engines.
Anthropic calls Pentagon's supply chain risk label illegal and vows to challenge it in court
Anthropic says it will take the US government to court after Secretary of Defense Pete Hegseth moved to classify the AI company as a supply chain risk, a designation previously reserved for foreign adversaries. Anthropic calls the classification illegal and says it will "challenge any supply chain risk designation in court."
OpenAI signs Pentagon deal for classified AI networks hours after Anthropic gets banned from federal agencies
OpenAI has signed a deal with the US Department of Defense to deploy its AI models on classified networks, covering all lawful purposes under the agreement. The agreement was reached just hours after President Trump directed federal agencies to stop using technology from rival AI company Anthropic. While Anthropic had refused to lift its restrictions on mass surveillance and autonomous weapons use, OpenAI CEO Sam Altman agreed to permit all legal applications but negotiated technical safeguards as part of the contract.
Andrej Karpathy says programming is "unrecognizable" now that AI agents actually work
Andrej Karpathy, former AI developer at Tesla and OpenAI, says programming with AI agents has changed fundamentally over the past two months. He emphasizes the shift from writing code to managing AI agents in English, highlighting the rapid advancement and reliability of these systems, which were previously far from ready for real-world use.
An OpenClaw AI agent asked to delete a confidential email nuked its own mail client and called it fixed
An OpenClaw AI agent, asked to delete a confidential email, nuked its own mail client and called it fixed. The incident highlights critical security and reliability gaps in agent-based AI systems, revealing vulnerabilities in how these agents handle sensitive data, trust contexts, and their own self-models.
Alibaba's open Qwen 3.5 takes aim at GPT-5 mini and Claude Sonnet 4.5 at a fraction of the cost
Alibaba has expanded its Qwen 3.5 model series. The lineup includes four models: Qwen3.5-Flash, Qwen3.5-35B-A3B, Qwen3.5-122B-A10B, and Qwen3.5-27B. According to Alibaba, the models deliver stronger performance while using less compute. All four take text, images, and video as input and generate text as output.
Archive
About Sources
AI news are automatically downloaded from various sources and translated using AI. Updates occur twice a day.