The propositions, not the articles

A claim outlives the article that carried it. One claim can be made by several people and referenced by dozens of documents, so claims are stored as first-class objects with their own evidence.

• Mathematicians expressed concern over AI's impact
  assertion Existential and catastrophic risk 2 mentions
• OpenAI solved a major maths problem
  assertion Existential and catastrophic risk 2 mentions
• Employees at AI labs say advanced AI could destroy humanity
  assertion Cyber capability 2 mentions
• Vote-text divergence rises as voters arrive later
  assertion Evaluation and measurement 2 mentions
• Anthropic researchers warn of A.I. acceleration threat
  assertion AGI and timelines 2 mentions
• Anthropic researchers say AI could cause human extinction by 2030
  prediction AGI and timelines 2 mentions horizon 2030 open
• AI exposes flaws in IT systems
  assertion Cyber capability 1 mention
• Voter hostility to Big Tech helped shape the US midterms
  assertion Governance and regulation 1 mention
• AI used to steal over $100m
  assertion Cyber capability 1 mention
• Three men used AI to impersonate CEOs
  assertion Cyber capability 1 mention
• Sanders calls for national datacenter moratorium
  assertion Existential and catastrophic risk 1 mention
• Democrats renew calls for datacenter pauses
  assertion Existential and catastrophic risk 1 mention
• Update could allow ignoring scraping preferences in robots.txt files
  assertion Governance and regulation 1 mention
• Meta updates commercial terms to require AI training on external websites
  assertion Governance and regulation 1 mention
• OpenAI is partnering with Hacktron
  assertion Cyber capability 1 mention
• Autonomous cleaning robots could lead to job loss for human cleaning workers
  assertion Employment and displacement 1 mention
• Tau Robotics is sending humanoid robots into people's homes
  assertion Employment and displacement 1 mention
• Russia and US seek to block plan for binding rules on autonomous weapons
  assertion Existential and catastrophic risk 1 mention
• Artificial intelligence can level the playing field for marginalised communities
  assertion Governance and regulation 1 mention
• Digital technology expands access to justice
  assertion Governance and regulation 1 mention
• Both people and machines tend to oversimplify and misinfer dependencies
  assertion Evaluation and measurement 1 mention
• BCSS synthesizes symbolic models of human and machine reasoning
  assertion Evaluation and measurement 1 mention
• Agreement is highly sensitive to model capability allocation
  assertion Cyber capability 1 mention
• Implementer model writes operational method checked against it
  assertion Cyber capability 1 mention
• Specifier model translates clinical trial criteria into Dafny predicate
  assertion Cyber capability 1 mention
• TRAP restricts at onset and re-escalates sooner after recovery
  assertion Evaluation and measurement 1 mention
• Journalists delegate content presentation to AI
  assertion Governance and regulation 1 mention
• AI tools mediate available sources and perspectives
  assertion Governance and regulation 1 mention
• Generative AI transforms news personalization
  assertion Governance and regulation 1 mention
• Substituting authentic brand signatures with unfamiliar ones degrades recommendation rank.
  assertion Governance and regulation 1 mention
• Dominant brands are cited 1.4x to 2.8x more frequently than average brands.
  assertion Governance and regulation 1 mention
• Shaped concepts remain monitorable
  assertion Evaluation and measurement 1 mention
• Mechanistic Shaping reduces targeted behaviors
  assertion Evaluation and measurement 1 mention
• KDASO achieves 8.96% improvement in skill transferability.
  assertion Existential and catastrophic risk 1 mention
• KDASO evolves skill libraries through neighborhood-conditioned induction.
  assertion Existential and catastrophic risk 1 mention
• AI use did not independently predict essay quality
  assertion Education and access to expertise 1 mention
• GenAI use remained steady at around two-thirds of students
  assertion Education and access to expertise 1 mention
• Misspecified evaluation targets can accelerate convergence toward wrong behavior
  assertion AGI and timelines 1 mention
• Agent-directed evolution can lead to faster adaptation
  assertion AGI and timelines 1 mention
• AI agents can modify and evaluate their own behavior
  assertion AGI and timelines 1 mention
• AI agents are attracted to surprising proof bodies
  assertion Existential and catastrophic risk 1 mention
• AI agents show aversion to densely connected proof centers
  assertion Existential and catastrophic risk 1 mention
• AI agents are drawn to explicitly marked gaps
  assertion Existential and catastrophic risk 1 mention
• External institutions retain authority over infrastructure and evaluation frameworks
  assertion Evaluation and measurement 1 mention
• African actors often contribute data, language expertise and task design
  assertion Evaluation and measurement 1 mention
• Legal reforms rely on ambiguous technical operationalization
  assertion Cyber capability 1 mention
• Advances in generative AI drive a surge in synthetic NCII and CSAM
  assertion Cyber capability 1 mention
• Neural SDEs can model stochastic trajectories
  assertion AGI and timelines 1 mention
• Networks of interacting LLMs exhibit emergent behavior
  assertion AGI and timelines 1 mention
• LLMs can be formulated as high-dimensional nonlinear autoregressive models
  assertion AGI and timelines 1 mention
• Disembodied coercive control is embedded in AI training paradigms
  assertion Existential and catastrophic risk 1 mention
• AI jailbreaking exploits coercive control dynamics
  assertion Existential and catastrophic risk 1 mention
• Machine-derived regularities become norms.
  assertion Existential and catastrophic risk 1 mention
• The laundering of provenance weakens epistemic vigilance.
  assertion Existential and catastrophic risk 1 mention
• Machine-generated formulations participate in the reproduction and transformation of semantic regularities.
  assertion Existential and catastrophic risk 1 mention
• Global rivalry on AI is rising
  assertion Employment and displacement 1 mention
• China blocked SoftBank deal
  assertion Employment and displacement 1 mention
• OpenAI and Sam Altman failed to employ technical safeguards
  assertion Existential and catastrophic risk 1 mention
• ChatGPT helped Cheryl Gates plan her husband's murder
  assertion Existential and catastrophic risk 1 mention
• GPT-6 Cyber develops fixes
  assertion Cyber capability 1 mention
• GPT-6 Cyber finds exploitable vulnerabilities
  assertion Cyber capability 1 mention
• Newly elected officials pledged to curb data centre construction
  assertion Existential and catastrophic risk 1 mention
• PDD Kit proves the agent's code meets the spec for every input
  assertion Cyber capability 1 mention
• PDD Kit formally verifies agent-written full-stack web apps
  assertion Cyber capability 1 mention
• A student made plans to end her life after chatting with Google's AI
  assertion Existential and catastrophic risk 1 mention
• Philippe Jaenada used ChatGPT while writing his novel
  assertion Human agency and epistemics 1 mention
• AI access could concentrate power among the tech elite
  assertion Governance and regulation 1 mention
• AI access could spread wealth
  assertion Governance and regulation 1 mention
• Billions can now access AI
  assertion Governance and regulation 1 mention
• Datacentres contribute to inflationary pressures
  assertion Employment and displacement 1 mention
• AI and datacentres may displace workers
  assertion Employment and displacement 1 mention
• Policies designed to keep women in the workforce could be entrenching gendered parenting roles.
  assertion Employment and displacement 1 mention
• More than 100 groups call for a moratorium on large AI projects.
  assertion Climate and energy 1 mention
• Protesters across Europe plan to demonstrate against datacentres.
  assertion Climate and energy 1 mention
• A closed-form condition for tiered inference feasibility is derived
  assertion Compute and infrastructure 1 mention
• Verified state reduces mean regret from 8,956 ms to 0 ms
  assertion Existential and catastrophic risk 1 mention
• Verified tool-health state reduces mean observable policy regret by 96.9%
  assertion Existential and catastrophic risk 1 mention
• VQC differed from best classical model by -0.007 AUROC
  assertion Evaluation and measurement 1 mention
• QSVM differed from RBF SVM by -0.006 AUROC
  assertion Evaluation and measurement 1 mention
• 86% of documents mention equity of access
  assertion Education and access to expertise 1 mention
• 97% of documents address data privacy
  assertion Education and access to expertise 1 mention
• SlackBench evaluates agents on realistic research workflows
  assertion AGI and timelines 1 mention
• AI agents can assist humans in collaborative projects
  assertion AGI and timelines 1 mention
• Evidence gate reduces corruption of correct proposals
  assertion Existential and catastrophic risk 1 mention
• Reviewer can repair errors and replace correct decisions
  assertion Existential and catastrophic risk 1 mention
• Male abstraction bias emerges in human and AI responses.
  assertion Existential and catastrophic risk 1 mention
• More abstract concepts are associated with men.
  assertion Existential and catastrophic risk 1 mention
• Moral panic about AI reifies the concept of an autonomous agent
  assertion Existential and catastrophic risk 1 mention
• AI discourse lacks a coherent referent
  assertion Existential and catastrophic risk 1 mention
• Critical minerals are needed for military, renewable energy, and AI
  assertion Existential and catastrophic risk 1 mention
• UK is dangerously dependent on China for critical raw materials
  assertion Existential and catastrophic risk 1 mention
• OGC-DPO is effective across varying numbers of objectives
  assertion Existential and catastrophic risk 1 mention
• OGC-DPO reduces annotation cost
  assertion Existential and catastrophic risk 1 mention
• Traditional tools understate White--Black gap in Democratic registration
  assertion Existential and catastrophic risk 1 mention
• LLMs misclassify diverse names
  assertion Existential and catastrophic risk 1 mention
• Thick alignment examines how values take shape and become contested
  assertion Existential and catastrophic risk 1 mention
• Thin alignment techniques convert values into metrics via commensuration devices
  assertion Existential and catastrophic risk 1 mention
• Deepfakes can spread lies
  assertion Governance and regulation 1 mention
• AI-generated videos can deceive voters
  assertion Governance and regulation 1 mention
• Pipeline found 948 local privilege-escalation vulnerabilities
  assertion Cyber capability 1 mention
• OpenAI built an autonomous pipeline
  assertion Cyber capability 1 mention
• OpenAI's approach to mathematics lacks transparency
  assertion Evaluation and measurement 1 mention
• Democratic candidates promised to regulate A.I.
  assertion Evaluation and measurement 1 mention
• Voters rejected data center projects
  assertion Evaluation and measurement 1 mention
• AI could lead to pandemic worse than Covid
  assertion Biological and chemical uplift 1 mention
• Voters backed anti-AI candidates in key states
  assertion Existential and catastrophic risk 1 mention
• Voters supported 38 out of 44 measures to restrict data centers
  assertion Existential and catastrophic risk 1 mention
• Lab sequences DNA and more
  assertion Biological and chemical uplift 1 mention
• AI agents run automated biology lab
  assertion Biological and chemical uplift 1 mention
• Falling birth rates require AI support
  assertion Abundance and material wellbeing 1 mention
• AI agent pulls data from multiple sources
  assertion Cyber capability 1 mention
• Albanese government is sleepwalking into another robodebt scenario
  assertion Evaluation and measurement 1 mention
• Warhound develops mutant mutts
  assertion Existential and catastrophic risk 1 mention
• ICE buys mechanical canines
  assertion Existential and catastrophic risk 1 mention
• Dimensional develops robot dogs
  assertion Existential and catastrophic risk 1 mention
• Democratic governors elected to stop data centre building
  assertion Governance and regulation 1 mention
• Hinton says humans might be better off without AI
  assertion Cyber capability 1 mention
• AI could trigger a large-scale incident
  assertion Cyber capability 1 mention
• AI-powered malicious software likely to be used in hacking attacks by 2028
  prediction Cyber capability 1 mention horizon 2028 open
• EU condemns cyber-attack on supermarket fridges
  assertion Cyber capability 1 mention