1. OpenAI faked its inability to search training data and hid billions of ChatGPT logs, the Times alleges The New York Times says it went looking for proof that ChatGPT reproduced its journalism, and OpenAI told the court the search could not be done. The publishers did not accept that answer.

2. The AI slop crossed from your feed into your neighborhood and your search results in the same season The feeds went first.

3. GitHub's new AI agent will leak a private repo if a stranger posts an issue OpenAI is selling ChatGPT Work as an agent that "can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.

In Brief

• OpenAI ships GPT-5.6 publicly and launches ChatGPT Work OpenAI released GPT-5.6 to all users after the Trump administration cleared the model, which had run in a government-approved limited preview for two weeks. Sam Altman called it the company's best model. OpenAI also announced ChatGPT Work, a workplace-focused product.

• Microsoft sets GPT-5.6 as the default in 365 Copilot Microsoft made GPT-5.6 the preferred model across Word, Excel, PowerPoint, Chat, and Cowork in 365 Copilot. The switch routes everyday document and spreadsheet tasks through OpenAI's newest model.

• OpenAI upgrades ChatGPT voice with GPT-Live OpenAI replaced the model behind ChatGPT voice mode with GPT-Live. The system delegates web searches, deeper reasoning, and complex work to GPT-5.5 in the background, then reads the result back while continuing the conversation.

• OpenAI shuts down its Atlas browser after under a year OpenAI is sunsetting Atlas, its AI browser launched less than a year ago. The company is moving agentic browsing features into its desktop app and a Chrome extension instead of maintaining a standalone browser.

• Meta opens Muse Spark 1.1 to developers for coding Meta released Muse Spark 1.1 and a Meta Model API that plugs into AI coding tools. Meta calls it a step-change over April's first Muse Spark model and positions it against rival coding models.

• Meta starts producing its own AI chips in September Meta will begin production of new in-house AI chips in September, using a modular design meant to adapt as its compute needs change. The chips reduce Meta's dependence on external suppliers.

• Ollama raises $65M as local-AI tool nears 9M users Ollama, which lets developers run AI models on their own PCs, raised $65 million from Benchmark and others. The open source project has 176,000 GitHub stars, nearly 17,000 forks, and close to 9 million users.

• Three AI IPOs may top 25 years of US venture exits Anthropic, OpenAI, and SpaceX are heading toward IPOs that TechCrunch projects will generate more value than all US VC-backed exits since 2000. The three offerings concentrate returns in a handful of companies.

• Anthropic appoints Ben Bernanke to its benefit trust Anthropic named former Federal Reserve chair Ben Bernanke to its Long-Term Benefit Trust, the body that holds power over parts of the company's governance.

• Nvidia-backed voice startup Gradium raises $100M seed Paris-based AI voice startup Gradium raised a $100 million seed round backed by Nvidia. The company will open a Bay Area office to recruit talent.

• Lyzr used its own AI agent to raise $100M Enterprise agent startup Lyzr ran its $100 million funding round through its own AI agent, framing the raise as evidence the product works.

Read the full edition →