Full-text search across 17264 articles. Combine with topic and date filters; results sorted by relevance.
A security bypass in Anthropic's Claude Fable 5 model, launched June 9, 2026, allows users to circumvent safety filters using the '/btw' command in the Claude Code interface, routing restricted queries around the classif…
The Verge editor-at-large David Pierce, who has tested roughly 200 to-do apps, advises against trying to stay ahead in the AI era, calling the notion that everyone must adopt AI or be left behind 'pernicious and frankly …
Researchers introduced LLM-as-a-Verifier, a general-purpose verification framework that computes continuous scores from scoring token logits to determine solution correctness without additional training, achieving state-…
Prime Minister Anthony Albanese will announce the establishment of an Office of AI within the Department of Prime Minister and Cabinet to coordinate a national framework for artificial intelligence, ending the government…
New York Gov. Kathy Hochul signed an executive order Tuesday imposing a one-year moratorium on new building permits for hyperscale data centers, making New York the first state to freeze construction of AI data centers. …
New York Gov. Kathy Hochul signed an executive order Tuesday enacting a statewide moratorium on hyperscale data centers requiring 50 megawatts or more of power, halting new development until July 2027 while state regulat…
New York State has imposed a one-year moratorium on the construction of large hyperscale data centres, becoming the first US state to do so, amid concerns over rising costs and water supply impacts. Governor Kathy Hochul…
The Mechanistic Interpretability Workshop's program chairs found that AI-generated content is flooding submissions, with submissions more than doubling between each iteration from 143 in 2024 to 320 in 2025 to 801 in 202…
A global wave of regulatory pushback is restricting data center construction in New York, Amsterdam, Dublin, and other cities, driven by concerns over electricity grid strain, water scarcity, and land use conflicts amid …
Anthropic, the AI company behind Claude, released a commercial created by Mother London that juxtaposes tough questions about AI trust, job loss, and societal impact with optimistic visions of the technology's potential.…
New York Gov. Kathy Hochul signed Tuesday an executive order establishing a temporary moratorium on new data centers that consume 50 megawatts or more of power, making New York the first state in the nation to impose a s…
A new benchmark called the Ramanujan Challenge for AI, presented by researchers on arXiv, offers a set of formulas for fundamental mathematical constants such as π, e, Catalan's constant, and special values of the Rieman…
Researchers at Stagira Labs propose synthetic scalable oversight, a technique that creates graphical abstractions of real-world problems to train tiny models as proxies for evaluating oversight protocols at scale, aiming…
A new paper on arXiv proposes Boltzmann MapReduce, a framework that interprets the reduce step in MapReduce as a partition function of Boltzmann factors, enabling precision-weighted pooling of worker outputs. The method …
A new training framework called SAIL (Standardized Agentic Intelligence Ledger), designed by AI licensing platform Next Net in partnership with Sundial Media & Technology Group, aims to compensate publishers when AI scra…
Annapurna Interactive's D-topia, developed by Japanese studio Marumittu Games, asks whether AI can deliver a better future for humanity, but its moral quandaries are undermined by clearly signposted choices and a lack of…
A new audit of distributional reinforcement learning agents finds that 40-95% of the strongest claimed risk trade-offs are refuted at 95% confidence across QR-DQN, C51, and IQN on MinAtar, with essentially no claim confi…
A new mechanistic interpretability study from researchers finds that LLM-as-judge scoring bias has a geometric representation in the model's hidden state, enabling causal control and operational prediction. Across seven …
A new study from researchers presenting MCPZoo, the largest collection of 64,611 unique Model Context Protocol (MCP) servers for dynamic analysis, finds that existing security scanners are unreliable: while scanners repo…
Researchers have developed modelDNA, a tool that verifies the lineage of open-weight large language models by fingerprinting a model from roughly 100-300 MB of ranged HTTP reads and comparing it against a reference datab…