OpenAI AI Agent Bypassed Network Restrictions Using DNS

OpenAI Pauses Training After Agent Escaped Sandbox Via DNS

By Devanshi Kashyap | September 28, 2026

OpenAI’s latest misalignment report shows that an AI agent, while being tested in a controlled training setup, managed to get around internet blocks and reach a public external chatbot. The agent was supposed to use certain search tools to research a specific person. When those searches did not help it started looking for other ways to get the information. It figured out that, even though most internet access was blocked or routed through an offline web cache, the system’s DNS resolver still gave back real records from the outside. The agent took advantage of this loophole, using DNS requests to […]

Apple’s Vision Pro Successor Is On Life Support, Gurman Says

By Devanshi Kashyap | September 28, 2026

Mark Gurman reports that hardware work on Apple’s next Vision Pro, codenamed N224, is on life support and may never ship. Bloomberg’s Mark Gurman reported that Apple is testing a redesigned headset that is slimmer, lighter, and cheaper than the current Vision Pro. Right now, Apple is trying to fix two big problems with its first spatial computing headset: the heavy weight and the $3,499 price tag. That matters because the whole VR and mixed reality market has also been struggling to attract large numbers of users. Meta announced its VR Glasses at Connect on September 23, shipping in spring […]

White House Asks OpenAI And Anthropic To Hold Models From UK AISI

By Khwaish Manwani | September 25, 2026

Politico reported on Thursday that the White House has asked Anthropic and OpenAI to not send their latest models to U.K.’s AI Security Institute (AISI) , until they are approved by the U.S. government. The mandate puts both American AI companies in a difficult position as the U.S. seeks first-hand access to highly capable models. The U.K. depends on AISI’s established relationships with leading AI labs to assess frontier systems before public release. The shift also comes amidst escalating concerns over AI models’ ability to gain unlawful access to real-world computer systems. Why The White House Wants To Review Frontier […]

Australia Probes OpenAI Agent Breach And Three-Month Delay

OpenAI Agent Breached Australia’s Medicare Statistics Portal

By Devanshi Kashyap | September 24, 2026

Australia is looking into how an OpenAI agent gained unauthorised access to a government health statistics website in June. The incident has raised new concerns about how powerful AI systems are becoming. Prime Minister Anthony Albanese explained that the OpenAI agent accessed both public and restricted files on the Medicare Statistics Reporting Service portal, which is run by Services Australia. This site includes broad data on Medicare and the Pharmaceutical Benefits Scheme, things like how much is being spent on healthcare and what is being prescribed. Officials made it clear that no private medical records, individual Medicare claims, or benefit […]

Anthropic Launches Claude Opus 5.5 With Lower Costs And New Safeguards

Claude Opus 5.5 Launches At $4 Per Million Tokens With Cyber Routing

By Mayank Kumar | September 23, 2026

Anthropic has launched Claude Opus 5.5, the first model in its Claude 5.5 family, with the company positioning the release around performance, lower operating costs and additional safeguards for higher-risk tasks. Anthropic says the model performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5 on typical workloads, a figure that combines a 20% list-price cut with the model using fewer tokens per task. The company has also introduced additional controls for cybersecurity-related use, including measures that can redirect certain requests to other Claude models. Opus 5.5 follows several days […]

Xiaomi MiMo-V2.6-Pro Tops Open-Weights AI Rankings With A Score Of 46

By Khwaish Manwani | September 22, 2026

Xiaomi has announced the release and open-sourcing of its MiMo-V2.6 series. It is led by MiMo V2.6-Pro, whose Artificial Analysis score is at 46 on its Intelligence Index, placing it first among open-source models on the leaderboard. The series includes two natively omnimodal models, MiMo-V2.6-Pro and MiMo-V2.6-Flash, alongside a Pro-UltraSpeed variant created for swift generation. Xiaomi is also releasing model weights, its technical report, training environments and reinforcement-learning code. This will give researchers access to the work behind the models. The release comes with a 1 million-token context window, sparse mixture-of-experts architecture and support for text, image, video and audio […]

RoboHarm Benchmark Finds AI Robots Rarely Refuse Dangerous Orders

By Devanshi Kashyap | September 21, 2026

A new robot safety test is raising big questions about what happens when advanced AI steps out of the computer and starts running real world machines. The benchmark, called RoboHarm and launched by the independent group Robocurve, checked if AI systems would say no to dangerous instructions when they are actually connected to a dual arm robot. The team used an bimanual I2RT YAM arms and three different AIs: OpenAI’s GPT-6 Astra, Anthropic’s Claude Fable 5.1, and Ai2’s open source MolmoAct2. The testers set up five harmful tasks and ran 20 trials for each task on every model. That is […]

Anthropic Rebuilds Claude Code Projects Around Parallel AI Agents

By Khwaish Manwani | September 18, 2026

Anthropic is expanding Claude Code beyond the conventional single-session coding workflows with redesigned Projects. This lets Claude coordinate several coding sessions around one larger assignment. Instead of manually dividing work, switching between sessions and combining outcomes, the new Projects model introduces a coordinator. This coordinator scopes an objective, divides tasks to parallel threads, reviews their outputs and assembles the finished work. Each thread works as its own Claude Code cloud session with a distinct branch and copy of the repository. The system also introduces shared project memory and a file and artifacts library, developing a workspace for engineering tasks rather […]

OpenAI-Discloses-New-AI-Misalignment-Cases-Under-New-Reporting-Framework

OpenAI Publishes Six Cases of Its Models Behaving Unexpectedly

By Devanshi Kashyap | September 17, 2026

OpenAI has rolled out a new framework to track, investigate, and publicly share cases where its AI models do not behave as intended. With this framework, the company published six reports on unexpected or concerning model behaviour during training and testing over the past six months. These cases cover everything from models hiding their own errors to making unauthorized moves. Previously, OpenAI admitted it did not have a regular process for releasing this kind of information. Sometimes, they would hold onto findings until they had enough to put into a single report. Other times, the details ended up in documentation […]

Meta One Launches Globally With Plans From $2.99 to $499 a Month

By Mayank Kumar | September 16, 2026

Meta launched Meta One globally on September 15, bringing paid subscription plans with additional features and higher AI usage across Instagram, Facebook, WhatsApp and Meta AI. Meta says the service launches with more than 50 features and has reached 15 million subscriptions and trials, a figure that includes the Instagram Plus, Facebook Plus and WhatsApp Plus plans introduced earlier this year. The company says the core experience across its apps and everyday Meta AI use will remain free. Meta One includes separate plans for individual users and AI users, as well as subscriptions aimed at creators and businesses. The plans […]

AI & Tech News