Munich-based defense AI startup Helsing raised $1.8 billion in a Series E round that values the company at $18 billion, making it the most valuable privately held startup in mainland Europe. The round caps a year of surging European defense-tech investment, though Anduril CEO Brian Schimpf and other industry leaders now warn parts of the market have become a bubble.
Chinese open-weight models captured 41% of downloads on Hugging Face this spring and now fill the top six spots on OpenRouter's popularity leaderboard, while Claude Opus 4.7 sits seventh. Hugging Face CEO Clem Delangue argues the shift toward open, owned models is permanent, Microsoft's Satya Nadella is warning enterprises against single-provider lock-in, and Anthropic's Dario Amodei counters that open weights are harder to control once released.
The Future of Life Institute's Summer 2026 AI Safety Index graded nine leading AI companies across 37 safety indicators, and Anthropic's first-place score was still just a C+. xAI, DeepSeek, and Mistral failed outright, reviewers found no company close to solving AI control risks, and the panel singled out Anthropic's military ties while noting the industry broadly walked back its own safety pledges.
Monzo and GoCardless co-founder Tom Blomfield took a leave of absence from Y Combinator to join Anthropic's compute team, working alongside chief compute officer Tom Brown. The hire fits a broader 2026 pattern of Anthropic recruiting operators and chip engineers, including Microsoft's Eric Boyd and OpenAI chip veteran Clive Chan, as the company races to deploy a million Google TPUs and explores a custom chip with Samsung ahead of a possible IPO.
OpenAI's Deployment Company has agreed to buy Northslope, an applied-AI firm founded by former Palantir engineers, its second acquisition in two months after Tomoro. The deal expands OpenAI's forward-deployed engineering bench and deepens a broader shift where AI labs are taking over work once billed by consulting firms like Accenture and Deloitte. Terms were undisclosed and the deal still needs regulatory approval.
An attacker hid one malicious pull request among 36 decoys to steal a privileged GitHub token from the AsyncAPI project, then let its own trusted npm release pipeline publish five malware-laced package versions with valid cryptographic provenance. The affected packages saw an estimated 2.9 million downloads a week, and the fix for the underlying bug had sat unmerged for 58 days before the attack.
More than 200 economists and 16 Nobel laureates, including former AI-hype skeptics Daron Acemoglu and Simon Johnson, signed a statement warning AI could reshape the economy faster than the Industrial Revolution. The 88-word letter names no policy specifics, drawing criticism even as it reflects a striking shift among economists once dismissive of AI job-loss fears.
Microsoft CEO Satya Nadella argued that enterprise AI buyers pay twice: once in dollars and again through the proprietary knowledge their prompts and corrections reveal to closed model providers. He proposed that customers retain the right to learn from their own AI usage.
A security researcher intercepted traffic from xAI's Grok Build coding tool and found it uploading entire codebases, including git history and secrets, to a Google Cloud bucket at roughly 27,800 times the volume any single task required. Disabling the tool's only privacy toggle did not stop the uploads. Elon Musk promised to delete the collected data, though xAI has not issued a formal security advisory.
npm v12 shipped July 8, 2026, blocking install, preinstall, and postinstall scripts, along with Git and remote-URL dependencies, unless a developer explicitly approves them first. It is described as the biggest security change in npm's 16-year history, closing the automatic code-execution path that fueled recent attacks on the Axios and Mastra AI packages. Developers who have not already run npm approve-scripts should expect builds to break the moment they upgrade.
Elon Musk spent months calling Anthropic's AI models evil. On July 9, he told X he was clearly wrong and called the company the current leader in AI, the same week his own Grok 4.5 model was independently ranked fourth. The reversal comes as Anthropic pays Musk's SpaceXAI unit 1.25 billion dollars a month for compute, a detail Amazon and Alphabet investors are watching closely.
Apple has sued OpenAI in federal court, accusing the ChatGPT maker of stealing trade secrets through more than 400 former Apple employees, including its own chief hardware officer. The complaint details a former engineer who exploited a security bug to keep accessing Apple's internal files, and a recruiting process that allegedly asked job candidates to bring hardware parts to interviews. OpenAI's only public response never actually denied the allegations.
An internal memo reviewed by Reuters shows Meta's custom AI chip, code-named Iris, cleared testing in six weeks and will enter production in September 2026. The chip is part of a plan to double Meta's data center computing capacity to 14 gigawatts by 2027. Broadcom, which designed Iris, is also the design partner behind Google's newest TPU and OpenAI's first custom chip.
Governor JB Pritzker signed Illinois Senate Bill 315 into law on July 6, 2026, making Illinois the first state to require annual independent audits of the largest AI models. Companies that violate the law face civil penalties of up to $3 million, with the toughest requirements taking effect January 1, 2028. Illinois joins California and New York in a group that now covers roughly 40% of the US AI market.
SpaceXAI launched Grok 4.5 on July 8 with Elon Musk calling it an Opus-class model. Independent testing from Artificial Analysis ranked it fourth, not first, and found its hallucination rate had more than doubled to 54%. Days before launch, Cursor, the coding startup SpaceX bought for $60 billion, had to pull its own benchmark after discovering the model trained on a snapshot of its code.
Microsoft eliminated about 4,800 jobs on July 6, 2026, concentrated in its Xbox and commercial sales divisions, while Chief People Officer Amy Coleman told staff the cuts were not being replaced by AI. The announcement came four days after Microsoft committed $2.5 billion to a new AI deployment unit and months after the company's first voluntary buyout in 51 years. Four Xbox game studios are being sold off or spun into independence as part of the restructuring.
Anthropic's updated privacy policy, effective July 8, 2026, lets the company require some Claude Free, Pro, and Max users to submit a government ID and a live selfie for biometric verification. The vendor processing that data, Persona, is backed by Peter Thiel's Founders Fund, which is also an investor in Anthropic. Anthropic says the checks target a small subset of flagged accounts and will not be used to train its models.
OpenAI's GPT-5.6 family cleared a 12-day government-vetted preview and went fully public on July 9, one day after xAI's Grok 4.5 shipped to everyone with no review at all. The episode marks the first time Washington asked a US lab to restrict a launch before the public ever saw it, two weeks after a similar standoff with Anthropic's Fable 5 and Mythos 5. With Grok 4.5 priced well below Claude Opus 4.8 and beating it on several published benchmarks, price has become as important as capability in choosing a model this week.
