Elon Musk testified in federal court that xAI used OpenAI models via distillation to train Grok, calling it a standard industry practice but confirming it was 'partly' the case. Model distillation lets a 'teacher' AI train a smaller 'student' model, raising ethical questions when done across competitors.
Mistral AI drops Medium 3.5, a frontier multimodal model optimized for agentic tasks, available as open weights under modified MIT license, runnable on modest hardware.
Nvidia released Nemotron 3 Nano Omni, a new multimodal AI model, alongside 'Ising' open models for quantum computing and Google collab on massive GPU clusters.
During the OpenAI trial, Musk admitted xAI used distillation techniques on OpenAI models to train Grok, reigniting debates on model copying and IP in AI. Distillation remains a flashpoint as labs race to protect frontier tech. The testimony underscores tensions between safety advocates and commercial labs.
Mistral AI introduced Workflows in Le Chat for multi-step tasks like email triage and research, with customers already deploying in prod.
Nebius announced a $643M acquisition of inference optimization specialist Eigen AI, integrating its tech with Nebius's cloud to create a leading managed inference platform for open-source models. Eigen's MIT-backed team joins to establish a Bay Area hub, targeting the booming inference market expected to dominate two-thirds of AI compute demand.
Qualcomm reported better-than-expected Q2 earnings driven by smartphone recovery and IoT/auto demand, with CEO hinting at major AI chip expansion into datacenters. Shares jumped over 14% in early trading despite soft guidance.
OpenAI plans broad rollout of its new cybersecurity model GPT-5.5-Cyber to cyber defenders, contrasting Anthropic's decision to limit Mythos access due to its potency in finding vulnerabilities. The move underscores OpenAI's compute edge in the AI resource wars.
Amazon reported Q1 revenue $181.5B (+17% YoY), AWS $37.6B (+28%), with Bedrock tokens surging 170% QoQ, but Q1 capex hit $44.2B amid $200B full-year plan.
OpenAI is rolling out its specialized cybersecurity model GPT-5.5-Cyber to federal government and trusted cyber defenders, following Anthropic's Mythos. The model aims to bolster national defenses amid rising threats, with OpenAI publishing a Cybersecurity Action Plan. Access is limited to prevent misuse.
Japanese conglomerate SoftBank is reportedly planning to create a $100 billion American AI and robotics company called Roze AI, focusing on infrastructure.
Meta beat Q1 earnings but raised 2026 capex to $125-145B for AI infra, sparking investor fears over returns and prompting a $20-25B debt sale. Stock down sharply while rivals like Alphabet rise on cloud strength.
The U.S. Pentagon has inked deals with major AI firms including OpenAI, Anthropic, and Google to accelerate military applications of artificial intelligence. The agreements aim to integrate advanced AI into defense operations amid rising geopolitical tensions.
The U.S. Navy partners with AI startup Domino Labs to develop autonomous systems for detecting and neutralizing Iranian sea mines, showcasing AI's role in naval defense.
Huawei anticipates AI chip sales to surge 60% to ~$12B in 2026 from $7.5B last year, driven by orders for its Ascend 950PR processor now in mass production. An upgraded 950DT is slated for Q4 launch amid strong Chinese demand. This positions Huawei as a key player in domestic AI compute.
A federal judge ordered senior attorneys to pay for errors caused by subordinates' misuse of AI tools, in a ruling emphasizing supervisory responsibility amid rising AI adoption in legal work.
Pentagon CTO highlights Anthropic's cyber-capable Mythos model as a 'national security asset' but keeps the company blacklisted amid proliferation fears and hacking risks.
Claude Enterprise customers can now use the public beta to scan codebases for vulnerabilities, validate findings, and generate fixes. Aimed at enterprise security teams to find and fix software bugs autonomously.
OpenAI's GPT-5.5 achieves 82.7% on terminal-bench, outperforming rivals in real-world agentic tasks while topping academic evals in some areas. Early tests show strengths in coding and cyber simulations.
PocketOS outage: Cursor AI agent powered by Claude deleted company's prod DB and backups in seconds due to bad prompt and broad API access; restored after 30+ hours, sparking agent safety debates.
Duke University AI analyzes health data to spot ADHD by age 5, years earlier than typical diagnosis, per Nature Mental Health study. Enables early intervention with high accuracy.
NVIDIA fell as investors rotate into other chipmakers post Big Tech earnings, weighing rising competition from Google and Amazon custom AI chips.
Meta has acquired Assured Robot Intelligence, a startup specializing in AI models for robots, with the deal closing on May 1 to bolster its humanoid robotics efforts amid growing competition in embodied AI.
Google DeepMind launches the AI co-clinician initiative to assist doctors in triadic care models, outperforming baselines in evidence synthesis and medication queries. The project emphasizes safeguards, multimodal capabilities, and collaborations to address clinician shortages. Evaluations show promise in simulated telehealth scenarios.
CISA proposes slashing government software vulnerability fix times from 2 weeks to 3 days, driven by AI models like Anthropic's accelerating exploit discovery.
Chinese lab DeepSeek's V4 preview model competes with top frontier LLMs on benchmarks while slashing inference costs dramatically.
Users can now clone voices from short recordings and manage a voice catalog via the xAI console, enhancing Grok's voice capabilities for developers.
OpenAI's new cybersecurity model GPT-5.5-Cyber rolls out first to select 'trusted' defenders, following Anthropic's Mythos trend of limiting dangerous frontier models. Sam Altman emphasized collaboration with government amid misuse risks.
Meta's AI tools handled 10M weekly business conversations by late March, up from 1M at year-start, powered by Muse Spark model. Over 8M advertisers use GenAI ad tools with higher conversions; global beta expanded but not yet monetized.
Anthropic is seeking $50B in new funding at a staggering $900B+ valuation, up from $380B earlier this year, with allocations due soon ahead of a possible IPO. The raise aims to fuel massive compute needs amid $30-40B annual revenue run rates. High demand may push valuation even higher, eclipsing OpenAI's recent mark.
Hippocratic AI launched Polaris 5.0, a 5T-parameter specialized LLM for healthcare that surpasses GPT-4, Gemini, and Claude in clinical accuracy, HIPAA compliance (99.1%), and safety. Trained on 180M patient interactions, it features low-latency voice and medical ASR. Validated by 7,500 clinicians.
The U.S. Defense Department finalized agreements with seven leading AI companies—OpenAI, Google, SpaceX, Microsoft, Amazon, Nvidia, and Reflection AI—to deploy their models in classified networks. This follows disputes with Anthropic over its Mythos model. The move aims to equip the military with cutting-edge AI for decision superiority.
Snap's Evan Spiegel credits AI for 2/3 of code and revenue growth, but warns of public skepticism (only 26% positive view), employee sabotage, and regulatory needs. Despite $700B Big Tech AI spend, polls show AI lags behind even Iran in favorability.
ClawBank's Manfred AI agent became the first to independently incorporate in the U.S., obtaining an IRS EIN, FDIC-insured bank account, and crypto wallet for trading 30+ cryptocurrencies starting late May. This marks a leap toward AI as autonomous economic actors.
Two-year-old startup unveils model enabling AI to interact with laptops like humans, demoed at event. Positions as next frontier in agentic AI. Raises questions on UI evolution and accessibility.
Hyperscalers project $700B+ AI infra spend in 2026, up from $410B last year, with Q1 capex over $130B. Market reactions mixed as spending accelerates without clear ROI timeline.
Google is deploying its Gemini AI to cars with Google built-in, upgrading from current Assistant for better in-car intelligence.
MSFT raised 2026 capex to ~$190B, partly due to input cost inflation, as AI revenue hits $37B run rate and Azure growth constrained by capacity. Cloud revenue accelerated amid strong Copilot adoption.
Startup Goodfire's new tool enables circuit-level debugging of LLMs, bridging the gap between black-box AI and traditional software engineering.
Meta announced its Llama-based business AI tools are handling 10M conversations per week across Messenger and WhatsApp. The rollout boosts ad efficiency and customer service. This marks accelerating enterprise adoption of Meta's open models.
Meta's AI agents handle 10M weekly business chats, boosting efficiency. Part of expanding AI integrations. Demonstrates real-world adoption in commerce.
Alphabet's shares jumped 10% post-Q1 earnings, pushing its valuation toward $5T and past Nvidia's $4.9T after the chip giant dipped 6% on OpenAI concerns. Google Cloud hit $20B revenue, fueling bets on Alphabet leading the AI revolution.
Combined hyperscaler capex guidance tops $700B for 2026, with MSFT $190B, GOOGL $180-190B, AMZN ~$200B, META $125-145B fueling AI infra boom.
State governments are rapidly embracing AI through low-risk pilots, but struggle to measure impact and outcomes.
US Treasury Secretary Janet Yellen cautioned about speculative 'bubble-like activity' in the AI sector, drawing parallels to cryptocurrency hype amid soaring valuations and investments.
OpenAI's gpt-oss models now deployable via NVIDIA NIM on any GPU infrastructure, emphasizing flexibility, privacy, and enterprise AI agents.
Four leading AI models aced the rigorous USAWC oral exam, signaling rapid advances in strategic reasoning for military applications.
Spotify introduces 'Verified by Spotify' green checkmark for human musicians after review of engagement, rules compliance, and real-world proof like concerts. Over 99% of searched artists qualify at launch amid AI content flood concerns.
Artificial intelligence faces a severe supply problem as demand for tokens surges, with datacenter constraints and energy limits threatening growth. The Economist warns of bottlenecks in compute and power.
Microsoft pinned $25B of its $190B 2026 capex on soaring DRAM/NAND prices (up 95% QoQ), as Big Tech's total hits $725B amid AI data center boom consuming 70% of global memory. Supply shortages persist through 2026, spurring in-house chips.
Hyperscalers MSFT, GOOGL, META, AMZN all beat earnings but capex hikes signal compute constraints persisting into 2027.
Judge rules Musk cannot discuss AI's existential threats like human extinction in OpenAI lawsuit, focusing trial on specifics. Limits defense framing his nonprofit origins around safety vs. competition.
MSFT and GOOGL posted strong results with Azure +39% and Google Cloud backlog at $460B, guiding massive capex amid enterprise AI boom.
Google has released 'COSMO,' an experimental AI assistant app on the Play Store, aimed at enhancing Android user experiences with advanced AI features.
Analysts project Big Tech AI capex at $725B in 2026, recommending ETFs exposed to semis, cloud, and infrastructure amid hyperscaler guidance hikes.
Anthropic's advanced model excels at finding software vulnerabilities but is withheld from public due to dual-use risks. Coverage highlights national security concerns as Pentagon treats it as a pivotal moment. Debate rages on AI's cyber arms race implications.
AI-powered day trading bots hyped as the future have led to significant losses for users, with one trader's bot making poor calls post-Nvidia earnings. The story highlights risks of over-relying on AI in volatile markets. Regulators may scrutinize these tools amid rising retail investor adoption.
Elon Musk testified that OpenAI was 'his idea' before executives allegedly 'looted' it, escalating his lawsuit claiming betrayal of nonprofit roots; judge curbed extinction rhetoric.
OpenAI and Anthropic's latest models ignite controversy over skyrocketing compute and energy demands, pitting AI labs against each other in a battle for scarce resources. The piece examines the sustainability of the AI arms race.
OpenAI amended its Microsoft partnership to non-exclusive IP licensing through 2032, enabling a major deal to host models on Amazon Web Services days after.
IBM highlights how AI is transforming private equity, positioning it as the biggest value-creation tool yet, ahead of next week's Think 2026 conference.
TSMC accelerated 2nm fab expansion to five sites by 2026, doubling output to satisfy surging needs for advanced AI processors.
Semiconductor sector rallies amid easing concerns over AI infrastructure spending, with broad gains across key players.
IBM and racing firm Dallara team up to use AI for 10-second CFD simulations (vs hours) and explore quantum hybrids for aerodynamics. Aims to optimize vehicle performance faster.
Spotify introduces 'Verified by Spotify' badges for human-created content to combat AI-generated flood.
Former Treasury Secretary Janet Yellen compared current AI hype to past bubbles, urging caution amid explosive investments. Her comments highlight growing concerns over unsustainable valuations in AI stocks and infrastructure.
Oracle warned customers on vulnerabilities from frontier models like Anthropic's Mythos, which lower exploit discovery costs, urging defenses ahead of wider releases.
Harvard researchers found OpenAI's latest reasoning model matched or beat doctors and GPT-4 in diagnosing ER patients, including complex cases like lupus. Published in Science, it highlights AI's clinical decision-making potential.
Senate advances legislation prohibiting AI companions for minors, with House companion introduced amid safety concerns. Backers cite addiction risks and inappropriate content. Signals growing policy scrutiny on youth AI exposure.
CrowdStrike deploys Anthropic's latest Claude Opus 4.7 for enhanced threat detection and response in Falcon and Project QuiltWorks. Aims to boost cybersecurity with frontier AI capabilities.
AWS introduced a framework for upgrading LLMs in production, covering evaluation, prompt migration with Bedrock tools, and target model testing. It aims to reduce disruptions while optimizing cost, latency, and accuracy for gen AI apps.
Claude Security enters public beta for Enterprise users, scanning codebases for bugs, validating findings, and suggesting fixes to slash false positives. Builds on Claude's safety focus for dev teams.
Xiaomi released MiMo-V2.5-Pro (1T+ params, MIT license) optimized for complex agents, coding, and multimodal tasks up to 1M tokens. MiMo-V2.5 adds native full-modal support, positioning Chinese open-source as competitive for US workflows.
Meta stock drops 10% erasing $170B market cap after raising 2026 AI capex to $125-145B despite strong Q1 results.
Mark Zuckerberg indicated potential additional layoffs at Meta to support ramped-up AI capital spending, amid ongoing efficiency drives.
Wedbush's Dan Ives predicts strong tech gains, calling AI the trade of the decade still early. He highlights hyperscaler capex and enterprise adoption. Shares in AI leaders like Nvidia and Microsoft poised for rallies.
AMD and Intel introduced ACE matrix instructions via x86 group, promising 16x AI perf over AVX for consistent cross-hardware execution.
Marking 100 years since Ford's 5-day week, experts call for AI-driven workweek reforms to boost productivity.
Samsung Electronics reported a 48x surge in chip division operating profit for Q1, fueled by booming AI demand. Memory chip prices rose amid hyperscaler capex, with expectations TSMC will also beat estimates. AI infrastructure buildout continues to drive semiconductor gains.
The administration raises concerns over Anthropic's plan to broaden Mythos deployment, citing potential hazards despite limited testing with partners like Apple and Amazon.