Latest AI (Artificial Intelligence) News
Anthropic raises $65B at $965B valuation, overtaking OpenAI
Anthropic reportedly closed a $65 billion Series H round, pushing its valuation to $965 billion and briefly surpassing OpenAI’s $852 billion mark. The round was led by major investors including Altimeter, Dragoneer, Greenoaks, and Sequoia, with participation from Google, Amazon, and semiconductor firms.
Claude Opus 4.8 launches with dynamic workflow automation
Anthropic shipped Claude Opus 4.8 with a major “Dynamic Workflow” upgrade for Claude Code. The model can plan complex tasks, spawn hundreds of parallel subagents, and verify its own outputs, with a reported 750,000-line codebase migration completed in 11 days at 99.8% test pass rate.
EU reaches provisional deal on AI Digital Omnibus
EU negotiators reached a provisional agreement on the Digital Omnibus on AI, which delays key enforcement deadlines for high-risk AI systems to 2027 and 2028. The package also adds targeted simplification measures and new prohibitions, signaling a softer but still expanding regulatory framework.
New Zealand issues guidance for AI regulation oversight
New Zealand’s Ministry for Regulation released “Responsible AI in Action” guidance for the country’s 267 regulators. The move is intended to help agencies apply AI responsibly as governments expand oversight of automated systems.
New Zealand warns of a surge in AI-driven cyber incidents
The New Zealand National Cyber Security Centre warned that frontier AI models could drive a “significant increase in vulnerabilities and incidents.” The warning reflects growing concern that advanced models are lowering the barrier for both discovering and exploiting software flaws.
Palo Alto Networks reports severe Claude-powered vulnerability hunting
Palo Alto Networks said Anthropic’s Claude Mythos model was able to detect and exploit software flaws at a level that generated two dozen security alerts in a single day. The company compared that to its usual rate of about five alerts per month, highlighting how AI is changing security workloads.
Tesla completes AI5 chip design tape-out
Tesla completed the design of its AI5 chip and reached tape-out stage, a key milestone before manufacturing. The development underscores the intensifying race among AI companies and hardware makers to secure custom silicon for training and inference.
OpenAI reasoning model reportedly proves 80-year-old number theory conjecture
OpenAI’s latest reasoning model is reported to have disproved Erdős problem 90, an 80-year-old conjecture in number theory. If confirmed, the result would be notable because it suggests AI systems are beginning to contribute to formal mathematics rather than only code and text generation.
Google expands AI Overviews with source labels and preferred publishers
Google is expanding Preferred Sources into AI Overviews and AI Mode, letting users highlight favored websites that receive visible “Preferred” badges in AI-generated answers. The company also introduced “Highly Cited” labels and article carousels to emphasize original reporting and trusted sources.
OpenAI tests richer ChatGPT ads and e-commerce features
OpenAI is testing upgraded ad formats inside ChatGPT, including larger images, customizable call-to-action buttons, and e-commerce layouts with pricing and reviews. It is also building more advanced ad infrastructure such as audience targeting, conversion tracking, and self-serve campaign tools.
DeepMind’s Hassabis says AGI timelines are shortening
Google DeepMind CEO Demis Hassabis warned that society may have only a few years to prepare for artificial general intelligence, with 2029 appearing plausible. His remarks add to the growing debate over how quickly frontier AI capabilities are approaching AGI-like performance.
DeepSeek cuts flagship AI model pricing by 75%
DeepSeek announced a permanent 75% price reduction for its V4-Pro AI model, sharply lowering API costs for developers and enterprises. The move increases pricing pressure across the global AI market, especially for vendors competing on inference and API usage.