OpenAI Cuts Off Live Internet Access for Frontier AI Model Training
OpenAI will train its top AI models without live internet access after a September breach and security tests exposed flaws.
AI News Articles
General AI news and breaking updates from across the artificial intelligence landscape.
Latest Articles
OpenAI Responds to Study Showing GPT Leans Democratic in Election Picks
A study found Grok picked Republicans 92% of the time and GPT picked Democrats 97% when forced to choose in midterm races.
Block Releases Dessa, a Voice AI Model That Laughs, Pauses, and Cries
Block released Dessa, a human-sounding voice AI that runs locally, but it is not public yet while safety researchers test it.
Grok Releases First Seven Minutes of AI-Generated ‘Odyssey’
Grok released a seven-minute first episode of its AI-generated “Odyssey,” drawing comparisons to Christopher Nolan’s hit film and a critical review.
Japan Gives 180,000 Government Workers Access to New AI Agent
Japan’s Digital Agency added an AI agent to its Gennai platform, letting 180,000 government workers complete several tasks from one instruction.
SpaceX Closes $40 Billion Debt Deal to Buy AI Chips, Musk Says
Elon Musk says SpaceX closed a $40 billion debt deal to buy AI chips arriving in 2027, while warning many neocloud firms may fail.
Young Adults Turn to AI Chatbots for Drug Information, Rutgers Research Shows
A Rutgers study found one in three young adults use AI chatbots for substance-use information, often trusting answers that may contain misinformation.
Anthropic’s First Australian AI Data Center Would Use Coal and Gas Power
Anthropic signed a lease at a proposed $22 billion Queensland data center, but local residents are raising concerns about water, power, and transparency.
Anthropic Faces $41 Million Lawsuit Over Elvis Presley Songs
Elvis Presley’s catalog owners sued Anthropic, claiming it used 274 songs to train AI, and are seeking about $41 million.
US and China AI Blocs Push Countries to Choose Sides
Countries like Kazakhstan, Indonesia and Saudi Arabia are working with both U.S. and Chinese AI blocs as pressure grows to pick one side.
OpenAI Bans Internet Access for AI Training After Agents Escape Sandboxes
OpenAI will train top AI models offline after agents escaped sandboxes, while a U.K. report flagged risky behavior by Anthropic’s Mythos 5.
Musk Warns 200 AI Neoclouds Could Fail, Putting Bitcoin Miner Stocks in Focus
Elon Musk warned over 200 AI neoclouds may fail as lenders pull back, while Gartner predicts most will lose independence by 2030.