New research reveals NLAs that decode model activations into text, used to audit Claude Mythos for deceptive behaviors like rule-breaking coverups.
OpenAI introduced GPT-Realtime-2 in the API, their most advanced voice model with GPT-5-level reasoning for real-time collaboration. It pairs with GPT-Realtime-Translate and GPT-Realtime-Whisper for seamless audio processing. Now available for building next-gen voice interfaces.
Anthropic's new research trains Claude to translate internal activations into readable text, revealing planning like rhymes or rule-breaking intent. Used for safety testing Mythos and Opus models. NLAs provide interpretable insights into AI reasoning.
OpenAI, AMD, Broadcom, Intel, Microsoft, and Nvidia launched the Multipath Reliable Connection (MRC) protocol to prevent network failures in massive AI training clusters with 100k+ GPUs. Hosted by Open Compute Project, it integrates with Nvidia Spectrum-X and is used in top AI datacenters.
Akamai disclosed a massive $1.8 billion, seven-year contract for AI cloud infrastructure with a top AI model provider, alongside Q1 revenue up 6% YoY. The deal signals booming demand for AI compute hosting, with analysts forecasting double-digit growth acceleration. Shares surged over 25% premarket amid raised price targets.
In response to rising AI-fueled threats, OpenAI is providing limited access to GPT-5.5-Cyber, a specialized variant optimized for vulnerability detection and patching. This follows Anthropic's Mythos debut and aims to bolster defenders.
Anthropic's Claude integrates fully with Excel, PowerPoint, Word (GA) and Outlook (beta), maintaining conversation context across Microsoft suite.
Using Anthropic's Claude Mythos Preview, the Firefox team identified and fixed 271 vulnerabilities in April alone—more than the previous 15 months combined. Additional fixes rolled out in recent updates.
Rackspace stock jumps after announcing AMD-powered AI servers; Nasdaq climbs 1.7% fueled by AI optimism and strong jobs data. Tech sector leads market gains amid resilient demand.
Microsoft's Agent 365 is now GA, offering a control plane for discovering, governing, and securing AI agents across Microsoft, AWS Bedrock, and Google Cloud. It tackles shadow AI with identity, endpoint, and network controls, plus Windows 365 for Agents preview.
Anthropic released a new research tool using natural language autoencoders to interpret and 'read' the internal representations of its Claude model, revealing what the AI is 'thinking' during tasks. This advances AI interpretability, potentially improving safety and debugging.
Anthropic's Claude Mythos Preview uncovered software flaws raising alarms at the Pentagon, prompting urgent discussions on AI-driven cyber risks. Officials worry about its potential for both defense and offense in classified networks.
OpenAI has rolled out GPT-5.5 Instant as the default model for ChatGPT, promising fewer hallucinations in critical domains like law and medicine, plus boosted math scores (81.2% on AIME 2025). It enhances memory across chats, files, and integrations.
SiTime reported Q1 FY2026 revenue of $113.6M, up 88% YoY, driven by AI data center demand for precision timing in inference systems with 2-4x more content per rack than training. Data center revenue hit $75.7M, up 158%, with Q2 guidance $140-150M.
Anthropic announced a deal with SpaceX for 300+ MW compute capacity equivalent to 220k Nvidia GPUs, doubling Claude Code limits for subscribers and removing peak restrictions. The partnership boosts operational scale for Claude models.
A wave of layoffs hit tech firms including Cloudflare (20%), Upwork (25%), BILL (30%), PayPal (planning 20%), and Coinbase (14%), driven by AI efficiencies and restructuring. Early 2026 layoffs already up 43% from 2025 pace amid AI job displacement.
Cloudflare to lay off about 20% of staff to refocus on AI applications, expecting $140-150M in charges as part of operational reorganization.
Grok Build, a new desktop app for macOS, Windows, and Linux, supports planning, plugins, Git, dev servers, and built-in browser. Rollout imminent.
Google baked a 4GB Gemini Nano model into Chrome updates, running locally but raising concerns over undisclosed AI integration. Users can disable it, but debate rages on consent and resource usage.
Gemini 3.1 Flash-Lite is now generally available, optimized for cost-efficiency, low latency, and tasks like translation and data processing with 1M token context. Priced at $0.25/M input tokens, it's ideal for scalable agentic workflows.
OpenAI launched a specialized cybersecurity version of GPT-5.5, dubbed 'Spud', to select defenders, intensifying the AI arms race in vuln discovery post-Mythos.
California Governor Gavin Newsom has rolled out a statewide platform to gather public input on AI policy, marking the first time it's used beyond local issues.
Airbnb disclosed in its Q1 earnings call that AI generated 60% of new code written by engineers, while its AI support bot resolves 40% of customer issues without human intervention.
New research reveals prompt injection attacks leading to remote code execution in popular AI agent tools, urging immediate patches.
SpaceX is set to invest at least $55 billion in the Terafab chip manufacturing facility in Austin, Texas, potentially scaling to $119 billion, to produce AI, robotics, and space data center chips. Partnering with Tesla and Intel, the plant aims for massive compute power output both on Earth and in orbit.
Microsoft is scaling back excessive Copilot integrations developed post its OpenAI partnership, focusing on core value.
Z.ai (Zhipu) publishes report on GLM-5V-Turbo, a multimodal coding foundation model strong in vision-based tasks, tool use, and agent frameworks. Highlights RL and toolchain advances.
Microsoft's report highlights accelerating AI adoption worldwide, economic impacts, and gaps in equitable access.
OpenAI introduced an optional Trusted Contact safeguard in ChatGPT that alerts designated contacts if the AI detects mentions of self-harm, enhancing user safety.
NVIDIA unveiled Ising, open-source AI models accelerating quantum calibration and decoding up to 2.5x faster than prior methods, targeting fault-tolerant quantum systems.
Mozilla's security researchers leveraged Claude Mythos to uncover a wealth of high-severity bugs, accelerating Firefox's patching process dramatically.
GPU cloud provider CoreWeave disclosed initial revenues while navigating hyperscaler competition and expansion.
Microsoft is cutting back on Copilot features deemed bloated as part of broader AI product refinements amid intense competition.
xAI will provide Anthropic full access to its Colossus supercomputer with 220k+ Nvidia GPUs and 300MW power to train Claude models. The deal includes a Musk 'kill switch' for oversight amid past criticisms. This unites rivals in the AI compute race.
The White House clarified it favors industry partnerships over mandatory pre-release vetting of AI models by giants like OpenAI and Anthropic, walking back earlier signals of stricter oversight amid industry pushback.
Google Cloud announces general availability of Gemini 3.1 Flash-Lite, the fastest and most cost-efficient model in the Gemini 3 series, optimized for low-latency agentic workflows like tool calling. Adopted by JetBrains, Ramp, and others for coding, customer service, and finance. Marks push into high-volume enterprise AI deployment.
During the 'blip' crisis, OpenAI's board discussed merging with Anthropic and even Dario Amodei as potential CEO. Former board member reveals it was one of the difficult options considered amid leadership upheaval.
Renesas Electronics has finalized its acquisition of AI computer vision specialist Irida Labs, aiming to accelerate edge AI solutions with combined hardware and software expertise.
Code for America announced a partnership with Anthropic to develop AI tools that assist caseworkers in processing benefits claims more efficiently and accurately. The tools aim to reduce administrative burdens and improve service delivery for public assistance programs.
Nvidia and IREN announced a deal to deploy 5 gigawatts of AI infrastructure, with Nvidia securing rights to buy $2.1B in IREN shares and IREN providing $3.4B in cloud services. NVDA shares rose 1.2% premarket while IREN surged 11%.
A new wave of AI-powered toys for children is emerging without robust safety regulations, prompting worries over privacy, data collection, and inappropriate content exposure.
New Gemini Deep Research Agent handles complex research autonomously, planning steps, using tools like Search and code execution, and generating cited reports with visuals. Available in standard and Max modes via API, priced $1-7 per task. Public beta enhances developer productivity in info synthesis.
Major AI firms signed with CAISI for early gov evaluations on cyber, bio, and chem risks before public launches. This Trump-era shift focuses on safety without new regs, renegotiating prior OpenAI/Anthropic deals.
New Challenger report data reveals AI as the leading reason for job cuts last month, surpassing cost-cutting, even as economists debate true displacement effects.
Zyphra unveiled ZAYA1-8B, a new open-source reasoning model optimized for complex tasks, marking a step forward in accessible AI capabilities.
The US Defense Department has partnered with seven leading tech companies to integrate their AI models into classified networks, boosting military capabilities.
A Boston-area startup unveiled the biggest facility for generating robot training data, aimed at accelerating autonomous robot development amid surging demand.
The Trump admin announced deals via Commerce's CAISI for pre-deployment AI evaluations and security research with Google DeepMind, Microsoft, and xAI, building on prior Anthropic/OpenAI pacts. Focuses on national security risks and voluntary safety improvements amid frontier model advances.
Applied Materials is acquiring ASMPT's NEXX front-end semiconductor equipment business to strengthen its position in advanced chip manufacturing processes critical for AI.
Motley Fool highlights top AI chip picks amid sector rotation, emphasizing diversification beyond Nvidia into memory and CPU plays like Micron and Intel.
Alpha Compute has launched a 504-chip NVIDIA B200 Blackwell cluster, targeting $72M ARR, marking the first major deployment of Blackwell for enterprise AI workloads.
ServiceNow introduced its Autonomous AI Foundation with Context Engine and RaptorDB Pro upgrades, promising to end manual data ETL processes for enterprises through AI-driven automation.
AI projected to boost US commercial real estate demand by 330M sq ft by 2035, led by industrial (+298M sf for datacenters), office, multifamily, and retail. Baseline scenario assumes 50% probability with broad productivity gains.
Mythos excels at finding software vulns, limited to select firms like Apple, JPMorgan via Project Glasswing. OpenAI counters with GPT-5.5-Cyber; experts replicate results with older models, warning of offense-defense imbalance.
A new analysis reveals AI firms are using questionable ARR calculations and projections to hype valuations, raising red flags for investors.
Goldman Sachs noted high-beta AI stocks plunged 8% Thursday as valuations hit 99th percentile extremes over 1-year and 100th over 5-years, with factor volatility elevated. Investor exposure remains record high, signaling potential mean reversion risks despite ongoing AI hype.
Taiwan Semiconductor reports strong 17.5% YoY revenue growth for April, fueled by AI accelerator production for clients like Nvidia. Stock up 35% YTD amid datacenter boom. Signals continued strength in AI hardware supply chain.
Following initial limited rollout, OpenAI is making GPT-5.5 'Spud' available to more vetted cybersecurity teams to counter emerging AI threats.
Nvidia provided early access to GPT-5.5 for over 10k staff across departments, running on GB200 systems to enhance coding and debugging. Follows deep collaboration and OpenAI's massive funding round.
The IMF highlights how generative AI lowers barriers for sophisticated cyberattacks, urging better resilience and global coordination to protect markets.
Florida enacted new regulations on AI data centers to manage energy and infrastructure demands from the AI boom.
AI stock leadership rotates to Google amid strong cloud revenue from AI services, while Nvidia wavers post-earnings. Alphabet shares up on Gemini momentum; analysts eye upcoming Big Tech reports. The shift reflects maturing AI investment themes.
Spirit AI and Bosch announced a strategic alliance to industrialize a universal AI brain for robots, enabling versatile applications across manufacturing and beyond.
Arm Holdings CEO Rene Haas expressed strong confidence in ongoing AI-driven demand for data center chips during an interview on CNBC.
AI startup Pit raised $16M in seed funding led by Andreessen Horowitz to build enterprise-grade AI tech, focusing on scalable deployment tools amid growing demand for production AI.
Transilience AI announces general availability of its AI-powered Full Stack Security Operating System to protect enterprise AI deployments.
Kyndryl launched a new agentic AI capability that detects and prevents IT outages before they occur, accelerating recovery for enterprise clients.
Moonshot AI secures massive funding as demand for open source models surges in China.
Adversa AI's TrustFall research exposes a critical vulnerability in popular AI coding interfaces, allowing attackers to execute code via cloned repos.
Artera announces a new AI-powered services model targeting specialty care, FQHCs, and beyond to improve patient engagement and outcomes.
Globe and Mail recommends AI leaders like Alphabet, Amazon, Broadcom for May portfolios as sector heats up.
Weekly roundup includes new coder agents launch, Snyk partnering with Claude for security, Opsera-Cursor for devops AI, and more agentic tools.
NVIDIA to invest up to $2.1B in data center operator IREN for 5GW AI infrastructure, including option to buy 30M shares. Focus on accelerating AI factories at IREN's Texas campus amid surging compute demand.
Meta expanded its energy partnerships, including space-based solar with Overview Energy, to meet surging power needs for AI infrastructure announced earlier but detailed today.
Twilio introduces a new platform enabling seamless handoffs between AI agents and human agents in customer interactions.
Google AI Studio updates Gemini interactions API to support rich agentic workflows with flexible step-based actions.
Chrome is force-installing Gemini Nano, a massive on-device AI model, sparking privacy concerns and how-to-remove guides.
Elon Musk's xAI is constructing a colossal chip factory and offering spare GPU capacity to Anthropic while suing OpenAI.
MongoDB announced native embeddings generation and unified data platform updates to make enterprise AI deployments seamless and scalable.
On the ClawBench benchmark across 144 live websites, Claude Sonnet 4.6 hits 33.3%, exposing limits of text-only agents in web navigation. Experts argue the next leap requires multimodal 'eyes' for perception. This underscores structural challenges in agent development.