Loading...
Please wait while we get things ready for you
Please wait while we get things ready for you
GPT-5.6 Sol, Terra, and Luna are officially here — but the US government decides who gets to touch them. OpenAI dropped its next-generation model series Friday in a limited preview accessible only to partners pre-approved by the Trump administration, marking the first time Washington has preemptively gatekept an American AI launch. The flagship Sol is OpenAI's most capable model yet, with agentic upgrades in coding, biology, and cybersecurity and a new "ultra" mode that...
The Trump administration has asked OpenAI to restrict the release of its next model, GPT-5.6, to a small set of government-approved partners before any wider rollout — marking the first time the U.S. government has preemptively stepped in to limit an American AI company's model launch. The White House's Office of the National Cyber Director and Office of Science and Technology Policy made the request as they build a new...
The AI security paradigm has flipped overnight. Governments are no longer reacting after releases — they're preemptively blocking them. Open-source models are arming attackers as fast as they're advancing capabilities. And model theft has moved from garage-level experiments to industrial-scale operations. The question isn't whether AI security is broken, but whether the current playbook can be fixed at all.
OpenAI just turned up the heat — and the spice. The company unveiled Jalapeño, its first custom AI inference chip, co-designed with Broadcom in a blistering nine months — a timeline that typically stretches into years for new processor architectures. Jalapeño is a "blank-slate" design purpose-built for large language model inference, not the general-purpose GPU approach Nvidia dominates. It's targeting "gigawatt scale" data centers from Microsoft and others starting...
OpenAI and Broadcom unveiled Jalapeño, their first custom AI inference chip — and the numbers demand attention. The processor went from schematics to fabrication readiness in nine months, a timeline that typically takes years. Early results show roughly 50% cost savings compared with standard AI GPUs (per Broadcom's CEO), and OpenAI's own models acted as a "virtual chip architect" to accelerate the tape-out, per president Greg Brockman. This isn't just another chip...
The AI landscape is in full chaos mode: Anthropic drops Claude Tag into Slack as its best models sit under government ban, OpenAI's GPT-5.5-Cyber cruises past the same regulatory tripwire without a scratch, China's GLM-5.2 eats Anthropic's lunch in Hong Kong, and the industry proves it can buy a primary election.
Anthropic launched Claude Tag on Tuesday — an always-on AI agent that lives inside Slack as a persistent team member. Type @Claude in any channel and it...
Qualcomm is in advanced talks to acquire AI chip startup Modular for roughly $4 billion, Bloomberg reports — but that number only tells part of the story. The deal, which could be announced within weeks, is just one data point in a staggering 48-hour cascade of capital flowing into AI infrastructure. Abu Dhabi's MGX raised close to $50 billion — one of the largest dedicated AI investment vehicles ever — to accelerate infrastructure and technology spending,...
The idea that frontier AI models capable of devastating cyberattacks are safely locked behind corporate walls is collapsing. The evidence suggests capabilities are already loose — and the genie isn't going back in the bottle.
The Five Eyes intelligence alliance — US, UK, Canada, Australia, and New Zealand — issued a rare joint statement Monday warning that advanced AI hacking models are "months away" from broad availability. The agencies say capabilities...
Intel and AMD — rivals who've spent decades battling for x86 supremacy — just did something unexpected: they teamed up. The two chip giants unveiled Advanced Compute Extensions (ACE), a new spec designed to make CPUs dramatically better at running AI models without needing a GPU. ACE targets matrix multiplication, the bedrock of AI inference, with dedicated hardware that delivers up to 16x more operations per instruction than the current AVX10 approach. It's not for training...
The US Commerce Department ordered Anthropic to seek federal approval before allowing foreign access to its Fable 5 and Mythos 5 models — the most aggressive government intervention into any AI company to date. The directive, issued weeks after Anthropic confidentially filed for an IPO valuing the company at $900 billion+, treats frontier AI models as controlled munitions. Trump later told Axios he doesn't view Anthropic as a national security threat, adding that CEO...
JPMorgan is pushing AI past the point of no return — from advisor to autonomous operator. A growing number of businesses, led by the largest US bank, are deploying Anthropic-powered AI systems that don't wait for human approval before acting. These agents execute transactions, manage workflows, purchase services, and interact directly with financial infrastructure, according to reporting from TheStreet. The old division of labor — AI advises, humans decide, humans act — is...
Senator Bernie Sanders introduced a bill Thursday to impose a one-time 50% stock tax on AI companies with over $200 million in annual revenue, creating a $7 trillion sovereign wealth fund that would pay every American a 5% annual dividend — estimated at over $1,000 per person. The proposal also requires companies to split their AI and non-AI businesses, with a new bipartisan commission wielding voting shares to block AI decisions "bad for the country." This is the...