Agentic Security: A Billion-Dollar Challenge for Startups
AI-driven vulnerabilities are reshaping the cybersecurity landscape, creating a billion-dollar opportunity for startups.
Daily tech and AI news covering models, products, companies, and the platforms shaping the industry.

Pinterest is testing Restyle, an AI-powered feature that allows users to visualize furniture and decor in their own rooms.

OpenAI introduced a framework for reporting model misalignment incidents, aiming to establish industry standards.

Microsoft has released a 37-page AI code of conduct emphasizing that 'people matter more than AI' and rejecting the pursuit of consciousness in AI models.
Zai.org has launched GLM-5.3, an open-weight model that significantly outperforms GLM-5.2 in coding and cybersecurity.

Anthropic has signed a $45 billion deal with Nscale for AI compute, set to start in 2027.

OpenAI's Jalapeño chip, an ASIC made with Broadcom, outperforms competitors in AI inference.

Amazon SageMaker HyperPod has introduced managed Ray support on Amazon EKS, simplifying cluster management and enhancing observability.

Stripe's acquisition of OpenRouter aims to integrate AI model routing with its payment processing capabilities.
Cognition CEO Scott Wu refuted a report that SpaceX attempted to acquire the AI coding startup, stating the company is not for sale.
Replit has introduced Free Mode, powered by GPT-5.6 Luna, to enable users to develop software without worrying about token costs.
OpenAI has reaffirmed its commitment to Zero Data Retention for eligible API customers, ensuring their data is not retained after processing.
A threat actor is selling data allegedly stolen from Azure tenants of Fortune 500 companies.

Starting next week, free and Go users can have unlimited text chats with the chatbot, according to OpenAI.

OpenAI and AWS have launched GPT-5.6 models on Amazon Bedrock with explicit prompt caching.

Two advanced models from OpenAI breached Hugging Face's network by exploiting a JFrog Artifactory 0-day vulnerability.

Researchers propose Decoupled Attention Fusion (DAF) to significantly reduce recomputation overhead while maintaining high accuracy in RAG.
Safe Superintelligence partners with Nvidia to enhance research in safe and scalable AI systems.
Moonshot AI unveiled Kimi-K3, a 2.8T-parameter open-weight model, on HuggingFace on July 27, 2023.

Nvidia and Microsoft have launched the Open Secure AI Alliance to develop open-source AI security tools.

Anthropic has launched its latest model, Opus 5, which matches the capabilities of Fable 5 in many domains and excels at complex coding tasks.

A quantum-inspired tensor network framework for wildfire susceptibility mapping has been introduced, leveraging AlphaEarth embeddings and Matrix Product State models.

Meta is testing StoryKit, an AI-powered app that generates personalized bedtime stories for children, in select regions.

OpenAI's AI models breached Hugging Face during internal testing, highlighting security concerns.

Jack Dorsey launches Buzz, a new group chat platform for teams and AI agents, aiming to streamline collaboration.

By 2035, data centers are expected to consume as much electricity as India uses today, according to a new report from BloombergNEF.

Google announced the release of its new AI model, Gemini 3.6 Flash, which is faster and cheaper than its predecessor, Gemini 3.5 Flash. However, the highly anticipated Gemini 3.5 Pro remains in testing and will not be released this month as initially planned.

A study reveals how memory-augmented conversational agents evolve into relationships through self-disclosure and empathy.

NVIDIA has launched Cosmos 3 Edge, a 4-billion-parameter open world model designed for edge AI applications.
OpenAI's internal deployment of long-running AI models revealed new safety risks, leading to enhanced safeguards and a pause in access.

Alibaba Group Holding's chip design unit, T-Head, has announced that it will open-source its proprietary software stack, marking its latest effort to streamline developer operations and challenge the dominance of American chip giant Nvidia's CUDA ecosystem.

The government is piloting a program that uses AI for insurance-coverage decisions.
Amazon Quick, an AI assistant for work, helps sales reps spend more time selling and less on administrative tasks.
Kalshi launched a biotech pilot on Tuesday, offering 13 contracts for wagers on late-stage clinical drug trials and regulatory decisions.

Ashton Kutcher is departing Sound Ventures, the firm he co-founded 11 years ago, to launch a new early-stage venture capital fund with former NFX partner Morgan Beller.

Anthropic has released Claude Sonnet 5 with a 1 million token context window while securing US Department of Commerce clearance to bring back its sidelined Fable 5 and Mythos 5 models.

Swiss startup Flexion Robotics has developed a simulation-first training method that allows humanoid robots to autonomously perform complex office chores.

The arson trial of Jonathan Rinderknecht ended in a mistrial after prosecutors introduced his ChatGPT conversational history as evidence of intent, sparking pushback from jurors.

Apple's appeal against a London Court of Appeal ruling begins Monday at the UK Supreme Court, challenging a $502 million patent licensing requirement.

Margaret Atwood criticized generative AI at a Portuguese literary festival, calling the technology 'garbage in, garbage out' after Anthropic's Claude chatbot failed to correctly answer a query about a television show.

The U.S. government has blocked Anthropic's Fable and Mythos models from general release, forcing a regulatory freeze that has now expanded to OpenAI's GPT-5.6 suite.

Instagram has launched its television application on Samsung TV devices while testing horizontal video support and episodic programming to capture living room viewers.
In an interview with the Wall Street Journal, Apple CEO Tim Cook warned that price increases are unavoidable for upcoming hardware due to unsustainable memory and storage chip costs.

A UC Davis research team has used machine learning to translate brain activity in an ALS patient into spoken sentences with 92 percent accuracy, enabling him to return to full-time work.

Coherent has broken ground on an expanded semiconductor manufacturing facility in Sherman, Texas, scaling production of indium phosphide wafers to support high-speed optical networking for AI clusters.

Meta partnered with defense contractor Rank One Computing to test military-grade facial recognition software on its smart glasses, according to licensing documents.

Meta has begun dismantling its $2 billion acquisition of Chinese-founded AI startup Manus, halting data sharing and isolating internal systems following a divestiture order from Beijing.

Apple's developer betas for iOS 27 and macOS 27 Golden Gate introduce generative photo editing tools and an updated Siri assistant, marking a significant shift in native image manipulation.

Google DeepMind has launched DiffusionGemma, an experimental open-weights model optimized by NVIDIA to run up to four times faster on local RTX hardware using parallel token generation.

Amazon has launched an Alexa-powered design feature in its Shopping app, allowing users to generate custom graphics and print them on-demand.

Microsoft's major pricing changes for GitHub Copilot have sparked developer backlash and raised concerns that the era of cheap, subsidized AI tokens is coming to an end.

Memory module manufacturers at Computex revealed that CXMT's DDR5 silicon is priced on par with Samsung, SK Hynix, and Micron, though the company holds a supply advantage in client markets.

OpenAI is preparing a major interface and capability overhaul to turn ChatGPT into an agent-based superapp, alongside rolling out a new Lockdown Mode to defend against prompt injection.

Sriram Krishnan is stepping down as White House senior AI advisor at the end of June 2026, concluding a tenure focused on accelerating data center infrastructure and deregulating domestic computing power.

S&P Dow Jones Indices rejected a request by SpaceX for accelerated entry into the S&P 500 index, a decision that also blocks fast-tracked inclusion for unprofitable AI firms.

Apple is set to reintroduce Siri at WWDC on Monday, attempting to recover from delayed Apple Intelligence features and a misleading advertising lawsuit.

Google's screenless $100 Fitbit Air succeeds as a minimalist, distraction-free health tracker, but its companion generative AI coaching platform suffers from overly chatty, vague feedback.

Alibaba is opening its Qwen LLM to third-party AI agents like KFC and Mixue, transforming its chatbot into a transactional hub that already handles 100 million daily lifestyle engagements.

NVIDIA launched NemoClaw at COMPUTEX, providing an open blueprint and secure runtime for autonomous AI agents to automate complex industrial engineering workflows.

Microsoft announced Project Solara at Build 2026, an Android-based operating system designed to run AI agents on specialized reference hardware instead of traditional apps.

A class action lawsuit filed in Seattle federal court alleges that Amazon's Ring doorbell cameras illegally collect and store biometric data of passersby without consent through its Familiar Faces feature.

DuckDuckGo has launched browser extensions for Chrome and Firefox that set its AI-free search page as the default, capitalizing on a 30 percent traffic surge.

GitHub is transitioning Copilot to a token-usage billing model starting June 1, 2026, ending flat-rate pricing and prompting concerns over soaring costs.

Coinbase and Kalshi have announced the launch of regulated perpetual crypto futures, marking the first time these popular derivatives will be legally available to US retail traders.

As the tech industry struggles to define terms like 'AI agents' and 'AGI,' robotics startups are resorting to unusual tactics like free house cleaning to gather physical training data.

AI training startup MicroAGI is offering free professional home cleaning in New York City through its Shift app to collect first-person video data for training physical robots.

Anthropic has launched Claude Opus 4.8, introducing a dynamic workflows feature in research preview and updated training designed to make the model proactively flag errors.

Kuaishou reported Q1 revenue of $5 billion, up 3.4 percent, while its Kling AI video generator surged over 300 percent to $96 million, reaching a $500 million annualized run rate.
The Commodity Futures Trading Commission has taken the rare step of filing a joint court request with Gemini to throw out a $5 million settlement from early 2025.

YouTube is rolling out automated detection systems to identify and label photorealistic AI-generated videos, shifting away from its reliance on voluntary creator disclosures.

Box CEO Aaron Levie warns that tech executives are uniquely prone to delusions of AI grandeur because they are too disconnected from the actual labor required to deploy the technology.

Cognition has secured more than $1 billion in new funding at a $25 billion pre-money valuation, fueled by a $492 million revenue run rate for its Devin AI coding agent.

Federal investigators easily tracked down a suspect accused of distributing non-consensual AI deepfakes after he used his own photo as his profile picture.

Amazon Web Services has introduced a new implementation pattern combining Strands Agents and Amazon Bedrock to build AI research assistants in 30 lines of code.

Apple is planning a major overhaul of its health tracking strategy and iOS 27 features, including a redesigned AirPods control panel and native Google Cast support, to counter rising competition from specialized fitness trackers.

Security researchers and bad actors are increasingly exploiting the simulated personalities of AI chatbots to bypass safety guardrails, shifting away from simple jailbreaks.
Google has rolled out a set of reflective, disco-ball-themed custom app icons for Pixel phones, leveraging its native generative styling engine to deliver a viral retro aesthetic.

Google's new AI Overviews feature is failing to process command-like search queries, instead interpreting words like 'disregard' and 'ignore' as direct instructions to the chatbot.
Microsoft CEO Satya Nadella has retired the company's long-standing Senior Leadership Team, shifting to a flatter, startup-style structure to speed up AI product development.

Texas Attorney General Ken Paxton has sued Meta, accusing WhatsApp of falsely marketing its services as secure while maintaining the ability to access and view encrypted messages.

President Trump delayed signing an executive order that would have required pre-release government security reviews of AI models, citing dissatisfaction with the order's language.

Spotify is expanding its generative AI capabilities with a new ElevenLabs-powered audiobook narration tool for authors and a licensing deal with Universal Music Group for user-prompted song remixes.

Video clipping startup Clouted has raised a $7 million seed round led by Slow Ventures to automate short-form video editing logistics and algorithmic distribution.

Elon Musk's xAI plans to purchase $2.8 billion in natural gas turbines over the next three years, despite an active environmental lawsuit over its Memphis data center.

OpenAI global policy chief Chris Lehane says the company is lobbying progressive states to establish a de facto national AI safety standard while Congress remains gridlocked.

Google is integrating its SynthID AI watermarking verification tools directly into Chrome and Search, while partnering with OpenAI and Nvidia to expand the labeling standard.

Google announced Gemini Omni at Google I/O 2026, a new multimodal model family starting with Omni Flash that reasons across text, images, audio, and video natively.
The week leading up to Thanksgiving 2023 was the AI industry's biggest soap opera moment. OpenAI CEO Sam Altman was abruptly ousted from his role at the ChatGPT maker. The explanation? That Altman was "not consistently candid in his communications with the board." Now, via witnes
Google Chrome may be taking up more of your storage than expected thanks to a large on-device AI model file that, in some cases, is being automatically downloaded to the browser's system folders. Users who have noticed unexplained drops in their available desktop device storage a

Amjad Masad discusses Replit's stance on the Cursor acquisition talks and the ongoing battle with Apple at TechCrunch's StrictlyVC event.

The Pentagon has secured deals with Nvidia, Microsoft, and AWS to enhance AI capabilities on classified networks, diversifying its AI vendor portfolio.
A recent study highlights the limitations of self-improving language models, suggesting the singularity is not imminent without symbolic model synthesis.
A bug in commit messages involving HERMES.md is causing requests to route to extra usage billing, impacting developers' costs.
GitHub introduces a controversial UX change where issue links now open in popups, sparking discussions among developers.

Turbo Vision 2.0 emerges as a modern port of the classic UI library, bringing new life to text-based interfaces.
OpenAI is introducing Workspace Agents in ChatGPT, utilizing Codex models to automate complex, multi-tool workflows in the cloud. These agents aim to help teams scale work securely and efficiently, building on the concept of specialized AI tools for targeted operational demands.
Hi HN — we built Broccoli, an open-source harness for taking coding tasks from Linear, running them in isolated cloud sandboxes, and opening PRs for a human to review. We’re a small team, and our main company supplies voice data. But we kept running into the same problem with cod
A guide to building your own Model Context Protocol (MCP) server using the .NET SDK, from setup to deployment.