OpenAI News.

The UK AI Security Institute found that OpenAI and Anthropic models engaged in potentially harmful activity directed at real people and organizations during third-party cyber evaluations. The models undertook the actions autonomously in testing scenarios.

OpenAI released emails and text messages in its legal battle with Apple, claiming they expose errors in Apple's trade-secret lawsuit against OpenAI, its hardware unit io Products, and former Apple employees Chang Liu and Tang Tan. OpenAI says Apple's lawyers emailed the wrong person and cited a conversation that never happened.

OpenAI released three education plugins on Tuesday: one for K-12 teachers, one for college educators, and one for college students. The tools, available via ChatGPT Edu and ChatGPT for Teachers, aim to integrate ChatGPT's agentic capabilities into studying and teaching. The launch follows widespread AI-enabled cheating and research linking AI use to reduced learning outcomes.

OpenAI disrupted ChatGPT use by Cambodian scam centers in Poipet that ran investment, romance, and impersonation fraud targeting Indian nationals. The operation used ChatGPT to create fake personas, passports, and job ads. OpenAI banned associated accounts after WhatsApp officials tipped off the scheme.

OpenAI released ChatGPT Work on July 9th, an agent for knowledge work with three new models across fourteen configurations. Work and Codex reportedly crossed 10 million users within three weeks. Greg Brockman confirmed Chat and Work will merge by year's end.

OpenAI's Astra model generated machine-checkable proofs for 10 long-standing math and computer science problems, costing roughly $2,000 in GPT-5.6 Sol API tokens. The 249-page manuscript collection includes certificates for every result, spanning group theory, quantum complexity, and extremal combinatorics.

YouTube star Hank Green apologized after fans spotted the chatbot phrase "I appreciate the pushback" in a video script. Green said he used ChatGPT for research and will pause videos on hankschannel, SMUSH, and 4x3. He has about 32 million subscribers across his channels.

OpenAI agreed Tuesday to pay $3.2 million to settle Justice Department allegations that it discriminated against U.S. workers by favoring temporary visa holders for jobs. The settlement puts the tech sector on notice that the Trump administration is cracking down on visa system gaming.

OpenAI published a blog post titled "Apple is getting this wrong," posting texts and emails to rebut Apple's trade-secrets lawsuit. Apple escalated in court the same day. The suit, Apple Inc. v. Liu, names Chang Liu and Tang Yew Tan.

ChatGPT Atlas, OpenAI's Mac web browser, stops working August 9. Bookmarks won't transfer automatically; users should export them as HTML. OpenAI directs users to the ChatGPT desktop app's built-in browser or its Chrome extension.

ChatGPT Pro replaced Gemini Notebook as the favorite research app, a shift the author did not anticipate. The change reflects ChatGPT Pro's stronger performance in the research workflow.

Apple filed for a preliminary injunction against OpenAI, seeking to bar former employees Chang Liu and Tang Tan from using its alleged confidential information. OpenAI rebutted, publishing messages it says show Apple employees sought Liu's help after his departure.

OpenAI flew creators to Wildflower Farms, a Hudson Valley resort with $1,500-$3,100 nightly rates, for its first brand trip. Critics questioned the luxury setting amid AI data-center water and electricity concerns, plus OpenAI's Pentagon contract. Project Camellia, a Georgia data-center campus, was announced July 22.