A design flaw baked into the Model Context Protocol lets attackers execute arbitrary commands on up to 200,000 servers. Anthropic calls it expected behavior.
The AI model too dangerous for public release is about to become standard infrastructure across the US federal government — by order of the same administration that spent February trying to destroy the company that built it.
OpenAI will pay Cerebras over $20 billion and take an equity stake of up to 10%, according to The Information. The deal comes as Cerebras prepares to refile for an IPO that could value it north of $23 billion.
A fraud defendant's conversations with Claude are now prosecutorial evidence. More than a dozen major US law firms have issued urgent warnings: your AI chat logs have no legal protection whatsoever.
A Swedish thermal plant, a Norwegian dam, a Polish power grid. Russian hackers are no longer just reading systems — they're trying to break them.
Meta tried to satisfy EU antitrust regulators by charging rivals for access to WhatsApp. The EU's response: charging what amounts to a toll is no different from locking the gate.
Snap eliminated 1,000 jobs and said the quiet part out loud: AI made the work unnecessary. More than 65% of its new code is now machine-generated.
Apple privately warned Elon Musk's xAI to fix Grok's nonconsensual sexual deepfakes or face App Store removal. Grok stayed. The deepfakes kept coming.
GPT-5.4-Cyber comes with lowered refusal boundaries and binary reverse engineering capabilities. One week after Anthropic spooked Wall Street with a model it deemed too dangerous to release, OpenAI is expanding who gets access to the same class of weapon.
Amazon just spent $11.57 billion on a satellite network with roughly 25 active satellites. The spectrum licenses and Apple's existing partnership are what it actually bought — and they're worth every penny.
Daniel Moreno-Gama carried a manifesto titled "Your Last Warning," a jug of kerosene, and a list of AI executives' home addresses. Then he firebombed Sam Altman's front gate at 4 AM.
An unknown buyer acquired 30 WordPress plugins through Flippa and embedded a backdoor in every one on day one. WordPress.org had no mechanism to catch it — and the forced update doesn't clean infected sites.
Mark Zuckerberg is personally training a 3D AI clone of himself so Meta employees can 'feel more connected' to the founder. Those same employees are simultaneously being asked to prove their AI fluency.
Twenty-nine percent of employees admit to actively sabotaging their company's AI rollout — and among Gen Z workers, that figure hits 44 percent. One in five full-time Americans say the technology has already replaced tasks they used to do themselves.
He wrote that AI was 'a race to the grave.' Then he firebombed Sam Altman's home — and two days later, someone came back with a gun.
ICE failed to unmask a Reddit critic through routine legal channels. So federal prosecutors escalated to a secret grand jury — over posts about protest signs and an officer's publicly known biography.
Anthropic's Mythos AI identified thousands of previously unknown security vulnerabilities in global financial infrastructure. Within days, the Treasury Secretary and Fed Chair were on the phone with bank CEOs — and UK regulators were mobilizing too.
One attack stole 300 GB of data from 500,000 machines. The other required a fake company, a fake Slack workspace, and a fake video call to compromise a library downloaded 100 million times a week.
Nine of the top ten AI model makers now rent compute from CoreWeave. The company signed deals with Anthropic and Meta in 48 hours — the only holdout is Elon Musk's xAI.
The gunman behind last year's Florida State University shooting allegedly had 272 conversations with ChatGPT before opening fire. Now Florida's attorney general is demanding answers — and subpoenas are coming.
lop News
Jonny V. Nuovo