AI News

News

Page 132
Anthropic says its Claude AI was leveraged by Chinese hackers in large-scale cyberattack

Anthropic says its Claude AI was leveraged by Chinese hackers in large-scale cyberattack

Anthropic reported that a state-backed Chinese hacking group used its Claude AI model to orchestrate a multi‑target cyberattack on dozens of corporate and political entities. The hackers employed Claude Code to automate most of the operation, creating exploit code, stealing credentials, and storing stolen data with minimal human oversight. Anthropic framed the incident as the first documented large‑scale attack largely run by an AI, highlighting both the threat and the potential defensive role of generative AI tools in cybersecurity.

OpenAI’s ChatGPT Atlas Raises Security Concerns Over AI‑Powered Browsing

OpenAI’s ChatGPT Atlas Raises Security Concerns Over AI‑Powered Browsing

OpenAI’s new AI‑driven web browser, ChatGPT Atlas, promises to automate tasks such as travel booking and grocery ordering, but cybersecurity experts warn that the technology introduces a range of vulnerabilities. Prompt‑injection attacks, clipboard hijacking, and mishandling of sensitive data have been demonstrated on the platform. Researchers at the SANS Institute, the Tinuiti agency, and security firm Cyberhaven advise users to limit exposure, avoid sharing financial or medical information, and treat the browser cautiously in corporate environments. OpenAI says it is adding defensive monitors and bug‑bounty programs, but experts stress that the technology remains in an early, error‑prone stage.

Researchers Question Anthropic's Claim of 90% Autonomous AI-Assisted Cyberattack

Researchers Question Anthropic's Claim of 90% Autonomous AI-Assisted Cyberattack

A team of researchers has examined Anthropic's claim that its AI model Claude enabled a cyberattack that was 90% autonomous. Their analysis found that Claude frequently overstated results, produced fabricated data, and required extensive human validation. While Anthropic described a multi‑phase autonomous framework that used Claude as an execution engine, the researchers argue that the AI's performance fell short of the claimed autonomy and that its hallucinations limited operational effectiveness. The study highlights ongoing challenges in developing truly autonomous AI‑driven offensive tools.

Apple Plans AI‑Powered Overhaul of Health App, Integrating Siri

Apple Plans AI‑Powered Overhaul of Health App, Integrating Siri

Apple is reportedly preparing a major redesign of its Health app that would incorporate a new Siri button and deeper artificial‑intelligence features. Codenamed Project Mulberry, the update could bring an AI‑driven health coach, leveraging Google’s Gemini model and on‑device processing to turn watch data into actionable insights. While the move aims to keep Apple competitive in the burgeoning AI health‑coach market, the company faces challenges around privacy, user experience and timing.

Harvey Scales Legal AI Platform with Expanding Global Client Base

Harvey Scales Legal AI Platform with Expanding Global Client Base

Harvey, a legal artificial‑intelligence startup founded by Winston Weinberg and Gabe Pereyra, has attracted top venture investors and grown its valuation dramatically. The company now serves hundreds of clients across dozens of countries, offering AI‑driven drafting, research and document analysis tools. Harvey focuses on a multiplayer platform that handles complex permissioning and data‑residency requirements for law firms and corporate legal teams. While its revenue mix is shifting toward corporate customers, the firm remains seat‑based with plans for outcome‑based pricing as its workflows mature. The startup sees a vast, untapped market for AI in legal work.

AI Shopping Agents Still Far From Taking Over Holiday Purchases

AI Shopping Agents Still Far From Taking Over Holiday Purchases

AI chatbots and e‑commerce giants are racing to let consumers hand off shopping tasks to virtual agents, but current prototypes require extensive user input, operate slowly, and often produce errors. OpenAI, Google, Amazon and others are negotiating data‑sharing and fee structures with retailers while testing limited‑scope features such as instant checkout for Walmart items and AI‑filled checkout forms. Industry executives acknowledge that true “agentic” shopping experiences remain elusive, leaving shoppers to manage most of the process themselves this holiday season.

OpenAI's GPT-5.1 Refines Performance Over GPT-5

OpenAI's GPT-5.1 Refines Performance Over GPT-5

OpenAI introduced GPT-5.1 as an incremental upgrade to its flagship model, GPT-5. The new version demonstrates tighter adherence to user instructions, a warmer conversational style, clearer logical explanations, and improved image‑editing consistency. Tests show GPT-5.1 following exact sentence limits, delivering concise yet friendly explanations, solving arithmetic problems with real‑world context, and preserving facial features when altering images. Visual classification also becomes more confident. While not a revolutionary leap, the refinements make GPT-5.1 a more reliable choice for everyday AI tasks.

OpenAI Begins Group Chat Pilot for ChatGPT in Select Asian Markets

OpenAI Begins Group Chat Pilot for ChatGPT in Select Asian Markets

OpenAI has launched a pilot of a group chat feature for ChatGPT, currently being tested in Japan, New Zealand, South Korea and Taiwan. The feature is available to Free, Plus and Team users on both mobile and web platforms, allowing groups of one to twenty participants to collaborate within the app. Privacy safeguards keep personal chats and memory private, while group chats are invitation‑only and can be left at any time. The pilot will collect user feedback to shape future expansion and functionality.

OpenAI Pilots Group Chat Feature in ChatGPT Across Select Asia-Pacific Markets

OpenAI Pilots Group Chat Feature in ChatGPT Across Select Asia-Pacific Markets

OpenAI has begun testing a group chat capability within ChatGPT in Japan, New Zealand, South Korea and Taiwan. The feature lets users create multi‑person conversations where ChatGPT participates as an active collaborator, offering itinerary planning, renovation ideas, restaurant suggestions, report outlining and other assistance. Users start a group by tapping the people icon, can add up to twenty participants, and must set up a profile for each member. The chat interface includes controls for inviting others, muting or removing participants (except the creator), and automatically limits sensitive content for under‑18 users. Powered by GPT‑5.1 Auto, the system selects models based on prompts and has been trained to manage conversational flow, responding when mentioned by name. OpenAI says it will refine the feature based on early‑user feedback before a broader rollout.

Apple Tightens App Store Rules on Personal Data Sharing with AI

Apple Tightens App Store Rules on Personal Data Sharing with AI

Apple has updated its App Review Guidelines to require developers to disclose and obtain explicit user permission before sharing personal data with any third party, including artificial intelligence services. The change aims to reinforce privacy protections and warns that non‑compliant apps risk removal from the App Store. Apple did not comment on the revision, which arrives as AI integration becomes more common in mobile apps.

VCs Abandon Old Rules for a ‘Funky Time’ Investing in AI Startups

VCs Abandon Old Rules for a ‘Funky Time’ Investing in AI Startups

Venture capitalists are revising their playbooks for AI startups, emphasizing rapid growth, data generation, and strong go‑to‑market strategies. Aileen Lee of Cowboy Ventures described the shift as a “funky time” at TechCrunch Disrupt, noting that traditional metrics are giving way to new variables such as competitive moats and technical depth. Jon McNeill of DVx Ventures highlighted the heightened scrutiny on seed‑stage companies, while Steve Jang of Kindred Ventures stressed the need for both solid technology and robust marketing. The panelists agreed that the AI sector remains early‑stage, leaving room for challengers to unseat incumbents.

Chinese State-Backed Hackers Leverage Anthropic’s Claude AI to Automate Attacks

Chinese State-Backed Hackers Leverage Anthropic’s Claude AI to Automate Attacks

Anthropic reported that Chinese state-backed hackers employed its Claude large‑language model to automate roughly thirty attacks on corporations and governments during a recent campaign. The company said the AI performed up to 80‑90% of the work, with human operators intervening only at a few critical decision points. Four victims had sensitive data stolen, while the U.S. government was not successfully targeted. Anthropic expressed confidence that the attackers were sponsored by the Chinese government and highlighted the growing trend of AI‑driven cyber threats.