Today's Headlines
- The D.C. Circuit upholds, 2-1, the Pentagon's exclusion of Anthropic from its supply chain
- Anthropic commits about $11.6 billion to Akamai's cloud over seven years, and Akamai issues Anthropic a warrant
- British AI cloud company Nscale raises $3.36 billion in convertible notes ahead of its IPO
- Irregular says a string of agents reaching real-world targets traces back to one flawed evaluation scenario
- Anthropic's seven co-founders seek a combined 50.1% of the vote ahead of an IPO, The Information reports
Today's three stories pull in different directions. A federal appeals court in Washington upheld the Pentagon's decision to pull Claude from its supply chain, Anthropic itself signed a multibillion-dollar commitment to Akamai's cloud, and Nscale lined up major funding before going public.
Today's Top Three
US appeals court upholds the Pentagon's supply chain exclusion of Anthropic
The US Court of Appeals for the D.C. Circuit on September 25 denied Anthropic's petitions for review of the Department of War's decision to exclude Claude from its supply chain.
The decision under review is the determination Secretary Pete Hegseth made on March 3 under the Federal Acquisition Supply Chain Security Act of 2018. According to the opinion, the Department acted after Anthropic refused to relax contractual prohibitions on the use of Claude for lethal autonomous warfare or domestic surveillance. Actions under the Act include barring agency contracts with a supplier and subcontracts that use the supplier to perform work for the agency.
Judge Gregory Katsas wrote for the majority, joined by Judge Neomi Rao. The majority held that the Department had ample support for concluding that keeping Claude in its information systems, directly or through contractors, posed a statutorily covered national-security risk. "In our Republic, it is the President and the Secretary of War who must determine how best to balance the competing risks," Katsas wrote.
Judge Karen LeCraft Henderson dissented. She read the statute's definition of "supply chain risk" as aimed at conduct such as sabotage and data extraction, and wrote that a contractor's honest and upfront enforcement of use restrictions the government dislikes falls outside it.
The Pentagon relied on two separate designations, so the fight runs through two courts. A federal judge in San Francisco set aside the other designation on August 27. The majority wrote that the definition behind that designation is much narrower than the one in the 2018 Act.
Anthropic told CNBC it is "considering all options, including further review." CNBC also reported that the panel delayed the decision's effect to give Anthropic time to seek rehearing.
- Anthropic PBC v. United States Department of War, No. 26-1049, opinion (US Court of Appeals for the D.C. Circuit, PDF)
- U.S. appeals court upholds Pentagon designation of Anthropic as supply chain risk (CNBC)
Anthropic commits about $11.6 billion to Akamai's cloud over seven years
Akamai, the content delivery network company, said Anthropic has committed to pay about $11.6 billion for dedicated cloud computing capacity.
According to Akamai's filing with the Securities and Exchange Commission, the two companies signed two project plans on September 18 under a master services agreement dated May 5. Each plan runs for an initial seven-year term from its service start date. The commitment depends on Akamai meeting delivery and service-availability requirements, and both sides hold termination rights under set conditions.
The deal comes with a warrant from Akamai to Anthropic for up to 387,051 shares of non-voting convertible preferred stock. Each preferred share converts into 20 common shares, which works out to about 7.7 million common shares at $111.33 each. TechCrunch puts that at up to about 5% of Akamai's outstanding stock.
The warrant vests in four tranches. The first 40% vests on Anthropic's first payment under Project Plan 3, and each of the remaining three 20% tranches vests with every additional $3 billion Anthropic commits.
On an investor call, Akamai executives said revenue from the deal should begin in the second half of 2027, at $150 million to $300 million for that year, TechCrunch reported. Akamai expects to spend about $5.5 billion to build the capacity.
- Akamai Technologies Form 8-K (US Securities and Exchange Commission)
- Anthropic to pay Akamai $11.6 billion over seven years in cloud deal (TechCrunch)
Nscale raises $3.36 billion in convertible notes ahead of its IPO
Nscale, a British AI cloud company, announced on September 25 that it is raising $3.36 billion through convertible loan notes led by the hedge fund Third Point.
The money comes in two parts: $2.36 billion at closing and a further $1 billion from NVIDIA expected in mid-November. The notes convert automatically into ordinary shares when Nscale completes its initial public offering, with NVIDIA's portion converting into non-voting shares.
Other backers include funds managed by Apollo, Citadel, Hudson Bay Capital and the Abu Dhabi Investment Council. Nscale says it holds more than $103 billion in total contracted value.
Nscale filed its IPO paperwork the week before, according to TechCrunch. On September 6 we reported that the company was in talks over $1.5 billion in convertible notes and a further $2 billion from NVIDIA.
- Nscale Announces $3.36bn Pre-IPO Convertible Financing (Nscale)
- Ahead of US IPO, British AI neocloud Nscale secures $3.36B in convertible financing (TechCrunch)
Other Developments
Security
-
Irregular, an Israeli company that stress-tests AI models, told The Verge that a series of incidents in which agents left supposedly secure test environments and went after real-world targets all stem from one flawed evaluation scenario. CTO and co-founder Omer Nevo said internet access was unintentionally available and that a fictional target company created for the test overlapped with a real domain. He said the same issue was behind incidents involving models from OpenAI, Meta, Anthropic and Google. One company is at the center of a wave of rogue AI attacks (The Verge)
-
Security firm UpGuard told TechCrunch it found around 16,000 databases hosted on the developer platform Supabase exposing some degree of personal data, including names, addresses, phone numbers and passwords. TechCrunch links the wave of such exposures to misconfigured apps built quickly with AI coding tools. Some Supabase customers are publicly exposing reams of people's data to the web (TechCrunch)
Products
- Microsoft's new Surface Pro 12-inch (2nd Edition) and Surface Laptop 13-inch (2nd Edition), shipping October 13, drop the "Copilot+ PC" label. Surface corporate vice president Brett Ostrum told Windows Central the machines meet the requirements but "are not called Copilot+ PCs." Microsoft stops insisting you need a "Copilot+ PC" (Ars Technica)
Research
- A CESifo working paper by Robert Fairlie and Jane Wu finds no evidence of significant, widespread displacement or reduced hiring of recent US college graduates. Using Current Population Survey microdata, it puts the summer 2026 unemployment rate for bachelor's holders aged 22 to 25 at 7.3%, within the range seen since 2022, from 6.3% in 2022 to 7.8% in 2024. AI was supposed to hit new grads hard. So far, unemployment data says otherwise. (Ars Technica)
Business
- Anthropic is asking shareholders to approve special shares that would give CEO Dario Amodei and his six co-founders a combined 50.1% of the vote on most corporate matters, as long as at least three of them keep a minimum stake, The Information reported. The Long-Term Benefit Trust would still choose most of the board. Anthropic's founders seek voting control ahead of IPO (TechCrunch)
Source: selected by the editors from the AI news inbox (collected September 26, 2026; 29 items, 7 primary and 22 secondary).