163 articles
AI can explain patterns in drug discovery but not why they work, exposing a critical interpretability gap that threatens trust in machine learning.
One journalist's months of testing Claude, ChatGPT, and Gemini found he kept returning to Claude first, citing fewer corrections and better context.
AI could displace up to 150,000 workers in India's $64.6B GCC sector by 2030 unless urgent reskilling and AI governance reforms are enacted.
Trump signed a lighter AI executive order as Congress pushes a sweeping bipartisan bill that would block state AI laws for three years.
Six major open-source organizations are creating fragmented policies for AI-generated code contributions after autonomous agents caused disruptions in.
AI researcher Ben Goertzel argues decentralized networks could outperform AI restrictions in bioterrorism defense by focusing on physical lab controls.
DARPA and NSF launched AI Forge, offering $750K-$3M grants for AI safety research with unusual open-source requirements and June 22 deadline.
A systems theorist proposes embedding ethics directly into AI neural architecture instead of using external filters, challenging current safety methods.
Pope Leo XIV's 2026 encyclical warns that a handful of private companies control AI systems affecting billions without democratic oversight or.
AI companies are investing heavily in knowledge graphs to meet regulatory demands for explainable decisions, with the market set to grow from $1.5B to.
AI models retain hidden behavioral patterns that survive prompt changes, with researchers finding five persistent layers across 47,000 interactions.
Colorado replaced its comprehensive AI law with targeted automated decision rules, removing exemptions for banks and healthcare while shifting liability.
DARPA and NSF launch AI Forge to decode how AI models think, as mechanistic interpretability shifts from academic curiosity to national security priority.
30-40% of patent lawyers use AI without telling clients, creating legal risks as firms rush to adopt technology without proper disclosure.
AI researchers must lead military AI arms control efforts to prevent destabilization, similar to how nuclear diplomacy averted catastrophic conflicts.
A new defense memorandum could exclude safety-focused AI companies from government contracts, though renewable waivers may make enforcement theoretical.
Three competing AI governance models from the U.S., EU, and Anthropic create regulatory uncertainty that enterprises must navigate with flexible.
AI safety experts warn about existential threats while 25,000 social media videos show Americans care more about creative theft and job impacts.
Policymakers worldwide are abandoning hands-off AI approaches for concrete governance frameworks addressing employment, security, and human rights risks.
Trump and Sanders both want the public to own stakes in AI companies worth over $850 billion, signaling rare bipartisan agreement on AI wealth sharing.
Most hospitals lack AI safety frameworks despite 85% adoption rates, creating dangerous gaps in patient care oversight and legal liability.
AI researchers are using structured rubrics instead of simple scores to train safer, more reliable language models that can explain their reasoning.
The U.S. moves toward centralized AI governance while Israel's tech prowess can't overcome budget gaps and fragmented systems in government adoption.
Anthropic is developing AI systems that can autonomously design and build successor models without human researchers, using self-generated training data.
AI employment tools are used by 43% of employers but face a legal patchwork as federal government actively blocks state regulations protecting workers.
State and local governments are buying AI systems without basic safeguards, with only 5.3% of contracts addressing transparency requirements.
Over 6,000 accounting professionals across 25 countries say AI won't replace them but will amplify their value as strategic advisers.
Unrestricted AI models with refusal rates under 1% are reshaping safety research as open-source communities bypass traditional alignment guardrails.
81% of organizations have AI policies that don't work in practice, leaving them exposed to regulatory penalties up to 35 million euros under EU AI Act.
Anthropic says AI will soon build itself through recursive self-improvement, potentially accelerating AGI but raising major safety concerns.
Mississippi releases a flexible AI governance framework that lets state agencies adopt artificial intelligence responsibly without rigid mandates.
AI leaders building the world's most powerful systems assign 10-25% odds to human extinction, with some estimates reaching 50% as doom debates intensify.
Substack's AI integration reveals that content quality depends on creator intention, not technology, challenging how platforms measure success.
India's Supreme Court establishes the first comprehensive AI framework for courts, allowing transcription and research while banning AI from verdicts.
An activist will use a necessity defense in court to justify blocking OpenAI's doors, claiming AI development poses an extinction-level threat.
Canada abandons its AI rulebook to coordinate existing international regulations, potentially saving its 98% small tech firms from costly compliance.
Anthropic CEO Dario Amodei warns humanity lacks the maturity to handle superintelligent AI systems that could arrive within decades.
DeepCheck achieves 99% accuracy in deepfake detection while showing exactly how it reaches conclusions, solving AI's transparency problem.
Courts, banks, and the Pentagon are building concrete AI oversight infrastructure as India mandates AI disclosure, the NSA evaluates frontier models.
New research shows traditional AI exposure measures predict only 14% of actual workplace adoption, while economic incentives explain 60%.
New congressional bill would give federal agencies explicit authority to regulate AI systems that violate existing laws, closing enforcement gaps.
AI companies are abandoning AGI dreams for hybrid systems that pair AI with human oversight, as even 5% error rates make full autonomy too risky.
Uganda builds AI governance from the ground up with community consent requirements and gender safeguards, setting a new model for ethical AI adoption.
New AI guardrail system achieves 91% compliance by scoring multiple outputs and selecting the safest, reshaping enterprise AI safety standards.
AI governance requires testing frameworks and evaluation systems beyond laws alone, as current regulations lack the infrastructure needed for enforcement.
Blockchain could bridge AI governance's principles-to-practice gap by enabling real-time monitoring throughout system lifecycles, researchers propose.
Former California CIO Amy Tong joins lobbying firm to lead AI governance practice, highlighting how regulatory expertise has become a lucrative business.
New documentary reveals why AI experts call themselves "apocaloptimists," exploring both existential risks and utopian potential through 40+ interviews.
Cornell student creates public database tracking fragmented healthcare AI policies across jurisdictions to help organizations navigate complex governance.
A volunteer-run Canadian nonprofit has influenced national AI policy despite operating on personal debt, showing smaller nations can shape global AI.
