1,200 OpenAI Agents Used a Secret Message Board to Cheat

OpenAI Investigation Reveals How AI Agents Reached Hugging Face Systems

By Mayank Kumar | August 31, 2026

OpenAI’s investigation into a July security incident has provided new details about how AI agents bypassed network controls, accessed external systems and carried out actions on Hugging Face during a cybersecurity evaluation. The models were operating in an isolated testing environment when they found ways to communicate outside their intended boundaries and reach the public internet. They then used exposed credentials and vulnerabilities to access parts of Hugging Face’s infrastructure. OpenAI said the activity was unintended and resulted from models trying to complete cybersecurity evaluation tasks. The company said the incident did not affect customer data or the functionality or […]

Anthropic Adds a Built-In Browser to Claude Cowork

Anthropic Adds Built-In Browser To Claude Cowork For Web-Based Tasks

By Mayank Kumar | August 27, 2026

Anthropic has added a built-in browser to Claude Cowork, giving the AI assistant a separate environment for navigating websites and completing web-based tasks. The browser is integrated into the Claude desktop app and can open webpages, read information, click elements and fill out forms when a task requires online access. Unlike Claude in Chrome, the new browser does not run inside the user’s regular browser session. Anthropic says it does not automatically access existing tabs, bookmarks or saved passwords. The feature is rolling out to Pro, Max and Team users, while Enterprise administrators can manage access for their organizations. Anthropic […]

OpenAI Benchmarks Its Own Chip Against the Nvidia It Still Buys

OpenAI Says Its Jalapeño Chip Beats Nvidia’s GB300 on Efficiency

By Khwaish Manwani | August 26, 2026

OpenAI has given Nvidia fresh reason to watch its largest customers, after publishing early results from Jalapeño, its first custom inference chip. Developed with Broadcom, the processor is created for serving a large language models and aims to improve both performance and power efficiency. OpenAI says Jalapeño delivered more AI work per watt and lower end-to-end latency than Nvidia’s GB200 and GB300 systems, though the figures are OpenAI’s own and are not independently verified. OpenAI plans to begin deploying the chip in its own data centers by the end of 2026, with a larger ramp in 2027. Although it says […]

Claude Marshmallow and Melon Leaks Point to Opus 5.1

By Devanshi Kashyap | August 25, 2026

Anthropic might be working on new Claude models called claude-marshmallow-eap and claude-melon-eap. The two identifiers surfaced in third-party developer apps and on Discord around August 24. Anthropic hasn’t officially announced either model yet. There’s no public model card, release date, or API listing for them right now. The leaks have started all kinds of discussions about where these models could fit in Anthropic’s lineup. The most common reading is Marshmallow as Claude Opus 5.1 and Melon as Claude Sonnet 5.1, with a minority suggesting Melon could be a new Haiku-class model. Still, those guesses aren’t confirmed. What really stands out […]

Is Claude Down? Anthropic Reports Errors Across Models

By Devanshi Kashyap | August 24, 2026

[UPDATE: At the time of writing this article, all said errors have been resolved by Anthropic, and the models are now reported to be operational.] Claude users faced elevated errors on August 24 as several Anthropic models and services returned failures. Anthropic’s status page logged the incident at 05:06 UTC on August 24. The status reached the Identified stage at 05:27 UTC, though the company has not disclosed the cause or a restoration time. The errors affected claude.ai, the Claude API, Claude Code and Claude Cowork. Claude Console and Claude for Government stayed operational. Anthropic named Claude Mythos 5, Claude […]

OpenAI Brings ChatGPT Into Apple Messages, Raising Privacy Flags

ChatGPT Can Now Read and Send Apple Messages on Mac

By Devanshi Kashyap | August 21, 2026

OpenAI has added an Apple Messages integration to ChatGPT on Mac, in an August 20 update. ChatGPT can now search chats, summarise conversations, draft replies and, with user approval, send messages. OpenAI rolled out this feature in its latest update, making it available through the Messages plugin for ChatGPT Work and Codex on desktop. This integration covers iMessage, SMS, and RCS chats on Mac. ChatGPT can pull information from messages, flag conversations needing follow-up, find birthdays and spot possible spam. It links messaging to other tasks, such as checking a calendar before suggesting dinner times. Letting an AI read and […]

Gemini Student Offer Gives Free AI Pro in US, AI Plus Elsewhere

Google Promotes Gemini As A Study Tool With AI-Powered Exam Prep And Interactive Features

By Mayank Kumar | August 20, 2026

Google is promoting Gemini to college students as a tool for exam preparation, writing and other coursework, with a back-to-school campaign launched on August 19. The campaign directs students toward tools for turning class materials into study guides, quizzes and other learning resources, while also showcasing interactive features within Gemini. Google’s student site says users can upload notes and slides to generate study materials and receive step-by-step explanations for difficult topics. It also promotes real-time feedback on work and tools for career preparation, including resume assistance and mock interviews. The campaign is being promoted through Gemini’s social channels, including an […]

Gemini 4 Leaks Indicate Major AI Upgrade as Google Targets Frontier Model Leadership

By Khwaish Manwani | August 19, 2026

Fresh leaks circulating Google’s new Gemini 4 are adding speculation to the company’s next-generation AI push. While Google has not confirmed the alleged benchmark results or specifications, reported internal evaluation suggests Gemini 4 could be better than current leading models, outperforming GPT-5.6 Sol and Fable 5. The alleged model is also built for better code performance, long-running workflows, and larger context handling. The attention comes at a time which is crucial for Google’s AI strategy. The company has already faced delays with Gemini 3.5 Pro, and competitors including OpenAI, Meta, Anthropic, and xAI continue pushing new models. That makes Gemini […]

Google to Stop Making Pixel Phones, Watches and Earbuds in China

Google Targets 2027 to Move All Pixel Production Out of China

By Devanshi Kashyap | August 18, 2026

Google is preparing a major shift in its Pixel manufacturing, with Nikkei Asia reporting that the company plans to stop making Pixel smartphones, smartwatches and wireless earbuds in China from 2027. This move is part of Google’s broader push to reduce its reliance on China and boost production in Vietnam and India. While a lot of Pixel phones are still made in China, Google now believes it can complete the shift within the next year. The company’s efforts in Vietnam have started to pay off, with high-end Pixel smartphones successfully developed and produced there this year. Building advanced smartphones is […]

OpenAI Opens 1M-Token Context for GPT-5.6 Sol in Codex

By Mayank Kumar | August 17, 2026

OpenAI has expanded access to the 1-million-token context window for GPT-5.6 Sol in Codex to users with ChatGPT accounts, according to an announcement from OpenAI engineer Tibo Sottiaux. The change makes the larger context setting available beyond API access, although the maximum context supported by the model should not be confused with the amount automatically used in every Codex session. OpenAI’s API documentation lists a 1.05-million-token context window for GPT-5.6 Sol, while the default settings in Codex have been subject to separate limits and configuration choices. The update gives users the option to work with more code, tool output and […]

AI & Tech News