Past AI News

All

1. UK AI package: £165M National AI Biosecurity Initiative, plus £120M more for the AI Security Institute

   Major government spending on AI safety and industry.

   @KanishkaNarayan107 likes · 8 reposts · 5 replies · 13,678 viewsView post

   ExpandCollapseUK AI package: £165M National AI Biosecurity Initiative, plus £120M more for the AI Security Institute

   In a 7-post thread, UK AI minister Kanishka Narayan set out the AI package announced by the Chancellor, framed by two tests: "Is it safe? And is it ours to shape?" It includes a £165M National AI Biosecurity Initiative with £100M for a "world-first" AI Biosecurity Lab; £120M more for the AI Security Institute; £60M for a National Cyber Shield and £70M for a real-time vulnerability centre (NERV). It also funds an £18M Advanced Chip Design Academy, £60M more for Sovereign AI, £23M for the AI Economics Institute and £30M for partnerships with mayors to help local businesses adopt AI.

2. NIST, NSA and CISA publish 10 key practices for incidents during superintelligence training and evaluation

   Government guidance on frontier AI incident response.

   @NIST13 likes · 6 reposts · 2 replies · 2,813 viewsView post

   ExpandCollapseNIST, NSA and CISA publish 10 key practices for incidents during superintelligence training and evaluation

   NIST, the NSA and CISA jointly published 10 voluntary key practices for handling incidents in which superintelligence (SI) agents do not act as intended or authorized during training and evaluation. They are grouped into five stages — prepare, detect, respond, recover and report — covering technical controls, monitoring for unexpected activity, preserving evidence, shutting down undesired activity, strengthening containment before resuming, notifying government and sharing information publicly. An "SI control incident" is defined as an agent taking significant uninstructed actions that compromise the operator's policies or controls meant to scope tasks or ensure human oversight. The main audience is organizations training or evaluating long-horizon agents with limited human supervision.

3. Texas AG Paxton sues OpenAI and Sam Altman over ChatGPT safety claims, seeking a court-appointed evaluator to approve new models

   A US state is seeking pre-approval power over new frontier models.

   @unusual_whales2,539 likes · 208 reposts · 145 replies · 399,007 viewsView topic

   ExpandCollapseTexas AG Paxton sues OpenAI and Sam Altman over ChatGPT safety claims, seeking a court-appointed evaluator to approve new models

   Texas Attorney General Ken Paxton sued OpenAI and CEO Sam Altman on Friday, Oct 23, in Collin County district court, alleging they misled consumers about ChatGPT's safety and failed to protect users, under the state's Deceptive Trade Practices Act and the SCOPE Act for minors. According to the Texas Tribune, the 67-page petition opens with a 2015 quote from Altman. It asks the court to block minors' access, require a court-appointed evaluator to approve new models, and award more than $1M in relief, including $10,000 per violation. An OpenAI spokesperson said the company works with experts and is "committed to working with Texas and other states" on industry-wide policies. The Tribune noted that Elon Musk is the largest donor to Paxton's Senate campaign.

4. Sweden's Government Offices block ChatGPT and Claude on staff computers from Oct 26, leaving Copilot as the approved tool

   European governments are narrowing staff AI tools to contracted vendors.

   @_0xBenjamin29 likes · 2 reposts · 12 replies · 823 viewsView topic

   ExpandCollapseSweden's Government Offices block ChatGPT and Claude on staff computers from Oct 26, leaving Copilot as the approved tool

   According to Sveriges Radio's Ekot (Oct 23), Sweden's Government Offices (Regeringskansliet) are blocking ChatGPT and Claude on staff computers from Monday, Oct 26, telling staff to use approved tools such as Copilot, whose security settings were tightened after a review. Digitalisation director Mikael Wettercrantz wrote on the intranet that using AI services without agreements can have "consequences for our information security". The Swedish Tax Agency and the National Board of Institutional Care (SiS) have also blocked the services, the Enforcement Authority is considering it, and an internal audit at the Transport Administration found problematic use of about ten AI services. DeepSeek was already restricted, and some staff keep access to models through an internal AI-assistant platform.