AI News: June 22, 2026

1. OpenAI Triples Revenue to $5.7 Billion in Q1 but Burns $3.7 Billion

OpenAI posted about $5.7 billion in first-quarter 2026 revenue while burning roughly $3.7 billion, with both figures tripling year over year, according to figures reported by The Decoder. Stock-based compensation alone accounted for more than $2.3 billion of the spend, and the company is said to hold around $73 billion in reserves, easing near-term pressure to raise fresh capital. The numbers show how fast OpenAI’s top line is scaling alongside equally steep cash consumption. Source

2. TechCrunch Weighs Who Benefits as US Cracks Down on Anthropic

Anthropic is at the center of a US government action that forced it to pull its Fable 5 and Mythos 5 models on national security grounds, and a TechCrunch Equity discussion examined what prompted the move and which rivals stand to gain. The episode frames the crackdown as a test of how export-control style restrictions land when applied to a leading domestic lab rather than a foreign one. It is the latest turn in a fast-moving policy fight over frontier model access. Source

3. Retailers Push to Exempt AI Ads From EU AI Act Deepfake Rules

Eurocommerce, a retail trade association whose members include Amazon, H&M, and IKEA, is lobbying to exempt AI-generated advertising from the EU AI Act’s transparency obligations. The group argues that an AI-generated product image, such as a synthetic living room staged to sell a sofa, should not be treated as a deepfake requiring disclosure. Zalando says roughly 90 percent of the marketing content on its platform is already AI-generated, underscoring how broadly the rules could bite. Source

4. UC Berkeley Study Ties ChatGPT to Grade Inflation in Writing and Coding Courses

UC Berkeley researchers analyzed more than 500,000 grades and found that courses heavy on writing and coding saw grades rise after ChatGPT launched. The effect concentrated in homework rather than exams, which the authors read as evidence that students are outsourcing work to AI rather than learning more. The findings add quantitative weight to concerns that generative tools are distorting academic assessment. Source

5. Data2Story Uses Seven Agents to Turn a CSV Into a Verified Article

Data2Story, a system from researchers at Oxford and Stanford, coordinates seven AI agents to convert a raw CSV file into a finished interactive news article with charts, web research, and source links, reporting that 93 percent of its statements were verifiable. In a reader study, 74 percent preferred the agent’s output to the original human version, though it only tied against elaborately produced long-form pieces. The work is a concrete demonstration of multi-agent pipelines applied to end-to-end data journalism. Source

6. TechCrunch Details Practical Apple Intelligence Features in iOS 27

Apple is shipping a set of Apple Intelligence features in iOS 27 that extend beyond the headline Siri overhaul shown at WWDC, according to a TechCrunch rundown. The piece highlights smaller, practical capabilities spread across the operating system rather than a single flagship assistant. It offers developers and users a clearer picture of where on-device and cloud AI will surface in everyday iPhone workflows. Source

7. Sam Altman Says a Generation of Researchers Underestimated Scaling

Sam Altman, OpenAI’s chief executive, used a Stanford talk to defend large language model scaling and argued that a generation of researchers slowed AI progress by underestimating what scaling could achieve. He pointed to OpenAI’s recent disproof of a mathematical conjecture as evidence that scaled models can produce genuinely novel results. The remarks restate OpenAI’s core bet at a moment when critics question whether scaling returns are flattening. Source

8. NYU’s Damodaran Warns an AI Crash Could Hit Harder Than Dot-Com

Aswath Damodaran, a finance professor at NYU, argued that a potential AI downturn could be more damaging than the dot-com collapse because the industry is pouring debt-financed capital into physical infrastructure rather than lightweight software. He added that even a technically successful buildout carries risk because the business case rests on replacing entire categories of jobs. The warning lands amid record data center and compute spending across the sector. Source

9. Signal’s Meredith Whittaker Warns AI Chatbots Are Not Your Friends

Signal president Meredith Whittaker publicly cautioned that AI chatbots are not friends, conscious beings, or sentient interlocutors, pushing back on the growing framing of assistants as companions. Her comments add a prominent privacy-focused voice to the debate over emotional reliance on AI systems. The intervention comes as several vendors lean into companionship and personal-assistant positioning for their chatbots. Source