AI Policy
Latest news, analysis, and updates on AI Policy from A47 News.
20 stories in Tech · Updated live
Trump Orders Federal Agencies to Replace 'AI' Terminology With 'Super Intelligence' in Official Documents
On October 27, 2026, President Trump announced via Truth Social that federal government documents will officially ban the term 'AI' in favor of 'super intelligence' or 'SI'. The announcement followed a September 2026 executive order and public criticism of media outlets such as The Wall Street Journal for continuing to use the original term amid rising public concern over job displacement, data center energy demands, and environmental effects. The directive applies only to executive branch communications and carries no authority over private companies or international bodies, so its practical reach and durability remain uncertain.
OpenAI Places GPT-6.1 Terra Into Public Arena Testing for Agent and Coding Evaluation
On October 26, 2026, OpenAI integrated GPT-6.1 Terra into Agent Arena and Code Arena for open public evaluation via Arena.ai platforms. The move follows internal use of Terra as a judge model and the September 29, 2026 public release of sibling model GPT-6.1 Sol, providing independent performance data beyond vendor self-assessment. Industry trajectory points toward greater reliance on third-party, task-based benchmarking as a standard for validating frontier AI agent and coding capabilities.
OpenAI and Anthropic Claim Refutation of Exponential Time Hypothesis Pending Independent Verification
On October 26-27, 2026, OpenAI and Anthropic each announced AI-generated claims that the Exponential Time Hypothesis is false, a foundational conjecture in computational complexity theory. The claims followed internal model work at OpenAI and were promptly reviewed by original ETH proposers Russell Impagliazzo and Ramamohan Paturi alongside Ryan Williams, though no independent verification or peer review has yet been completed. If verified, the results would undermine hundreds of conditional hardness results built on ETH and could reopen pathways to faster algorithms and reshape assumptions in theoretical computer science.
Netherlands Plans to Eliminate 25,000 Central Government Positions by 2035 Through AI-Driven Efficiency Drive
On October 26, 2026, the Dutch government announced a plan to reduce its central civil service by approximately 25,000 full-time equivalent positions by 2035, targeting €3.5 billion in annual savings through expanded AI deployment and other efficiency measures. The move follows a 41 percent rise in central government staffing since 2018, reaching a record 160,000 employees by end-2025 despite prior pledges to shrink the civil service and mounting fiscal pressure. The cabinet will present specific AI productivity targets and savings projections in Q1 2027, with cuts phased in gradually across ministries and agencies while police and defense are exempt.
Latest Stories
Trump Orders Federal Agencies to Replace 'AI' Terminology With 'Super Intelligence' in Official Documents
On October 27, 2026, President Trump announced via Truth Social that federal government documents will officially ban the term 'AI' in favor of 'super intelligence' or 'SI'. The announcement followed a September 2026 executive order and public criticism of media outlets such as The Wall Street Journal for continuing to use the original term amid rising public concern over job displacement, data center energy demands, and environmental effects. The directive applies only to executive branch communications and carries no authority over private companies or international bodies, so its practical reach and durability remain uncertain.
OpenAI and Anthropic Claim Refutation of Exponential Time Hypothesis Pending Independent Verification
On October 26-27, 2026, OpenAI and Anthropic each announced AI-generated claims that the Exponential Time Hypothesis is false, a foundational conjecture in computational complexity theory. The claims followed internal model work at OpenAI and were promptly reviewed by original ETH proposers Russell Impagliazzo and Ramamohan Paturi alongside Ryan Williams, though no independent verification or peer review has yet been completed. If verified, the results would undermine hundreds of conditional hardness results built on ETH and could reopen pathways to faster algorithms and reshape assumptions in theoretical computer science.
OpenAI and Anthropic Models Claim Refutation of Exponential Time Hypothesis Under Expert Review
On 26 October 2026 OpenAI announced that its internal AI models had produced a claimed refutation of the Exponential Time Hypothesis, a foundational computer science assumption proposed in the mid-1990s, with Anthropic releasing a similar claim hours later. The claims, partially verified in Lean and now under independent review by an expert working group including ETH originators, assert a sub-exponential time algorithm for counting satisfying assignments of fixed-width CNF formulas. If confirmed, the result would require revisions to the hundreds of algorithms and complexity papers that rely on ETH for conditional hardness results, marking a significant precedent for AI-driven scientific discovery.
Sweden Government Offices Block ChatGPT and Claude Web Access on Work Computers Over Data Security Risks
Effective October 26, 2026, Sweden's Government Offices prohibited browser access to ChatGPT and Claude on staff computers, affecting roughly 5,100 employees supporting the prime minister and ministries. The restriction was triggered by data security concerns over sensitive information being exposed to AI services without formal data protection agreements, following an internal review at the Swedish Transport Administration that flagged problematic AI usage with confidential data. The move is expected to push government AI use toward contractually governed tools such as Microsoft Copilot and domestically hosted platforms, reinforcing a broader European shift toward enterprise data controls over consumer AI services.