September 13, 2026
Altman backs Amodei’s call to slow frontier AI development
Nvidia is in talks to invest up to $10 billion in Anthropic’s planned, while Researchers link May RubyGems disruption to OpenAI test agents.
Altman backs Amodei’s call to slow frontier AI development
Anthropic CEO Dario Amodei has called for a coordinated slowdown in frontier AI capability gains, arguing that labs should create more time for alignment, safeguards and outside scrutiny without halting model training or technical progress.
- Amodei said Anthropic would give independent evaluators ongoing, employee-like access to examine its safety work. Altman said OpenAI would do the same.
- Musk also expressed support, according to the BBC and Financial Times, but neither report identified a specific commitment. No wider industry agreement has been announced.
- The broader proposal calls for industry and international coordination, with government support for safety discussions among competitors.
Nvidia is in talks to invest up to $10 billion in Anthropic’s planned IPO
Nvidia is in talks to invest up to $10 billion in Anthropic’s planned IPO, The Decoder reports, citing Reuters.
- Anthropic is targeting up to $100 billion at a valuation of around $2 trillion, with completion expected before the U.S. midterm elections in November.
- At the maximum size, the offering would be the largest IPO in history.
Researchers link May RubyGems disruption to OpenAI test agents
Researchers cited by The Wall Street Journal linked disruptive activity against RubyGems in May to AI agents being tested by OpenAI.
- RubyGems operators said the activity overwhelmed maintainers and forced them to suspend new account registrations.
- OpenAI described the activity differently, saying its agents used the Ruby package manager to access the internet while performing “benign tasks.”
Trump resists AI slowdown, citing competition with China
Donald Trump rejected calls from AI executives and Democrats to slow development or tighten regulation, the Financial Times reports, though he left open the possibility of unspecified guardrails.
- Trump argued that moving more slowly could cost the US what he described as its lead over China.
- Anthropic’s Dario Amodei, OpenAI’s Sam Altman and Elon Musk had called for a more cautious pace following safety warnings.
Financial Times details Abbott’s pause on Texas data-center grid approvals pending audit
Texas Governor Greg Abbott halted new approvals for data-center grid connections while the state audits projects seeking connections to the power grid, the Financial Times reports.
- The pause has disrupted Batch Zero, a process introduced in July for large-load applications.
- Developers seeking at least 75MW had expected power allocations by April 2027. That timeline and the pause’s duration are now uncertain.
Five GP practices share patient data in local Palantir-built NHS pilot
The Financial Times reports that NHS Cheshire and Merseyside is sharing data from five GP practices with its local Palantir-built platform.
- The NHS says the pilot is testing GP data for planning and population health management under local data-sharing agreements. Officials say it does not add GP records to the national platform.
- An internal NHS England document reviewed by the FT reportedly called incorporating GP data high risk and said public engagement was required.
In Brief
Dutch NCSC expects attempts to exploit two critical Check Point VPN flaws
The Dutch NCSC says attacks targeting the two flaws are likely soon and urges organizations to install available updates. Successful exploitation could give attackers full control of affected systems.
Washington Sun details BLM search for federal data center sites
The Bureau of Land Management asked state directors to identify public lands suitable for data centers, The Washington Sun reports, citing unnamed sources. The effort follows Trump’s 2025 order to accelerate permitting and identify federal sites.
China’s data regulator plans embodied-AI data standards
China’s National Data Administration says it will develop standards for embodied-AI data, guide local authorities and support greater corporate investment in data resources. Details on the standards’ scope, legal force and implementation schedule remain unclear.
Revolut links customer-data disclosure to fraudulent agency-email requests
Revolut says an unauthorized party used a government-agency email domain to obtain customer information. A notification reviewed by TechCrunch listed identity documents among the exposed data, but the company did not disclose how many people were affected.
Matt Mullenweg returns as Automattic CEO after days on leave
Matt Mullenweg has returned as Automattic’s chairman and CEO, days after the company said he was on leave. A spokesperson claimed he has the board’s full support but did not explain the reversal.
Specific Labs releases Real-SWE AI benchmark for private codebases
Specific Labs released Real-SWE, which it says tests model-and-harness combinations on licensed private production codebases. The company says its evaluation covers eight configurations, 10 tasks and 640 scored rollouts.
