1. Anthropic Spent the Week Courting Seoul. The White House Made It Cut Off One of Its Biggest Korean Customers. Anthropic opened a Seoul office this week and named a roster of Korean partners, its formal entry into a market it had been circling for a year.

2. Researchers call hacking-capable AI inevitable as the Pentagon puts generative tools in 1.5 million hands AI models that can hack at an advanced level will soon be ordinary, according to an Ars Technica report arguing the capability arrives regardless of what defenders want.

3. Adobe just made a chatbot the front door to Photoshop and Premiere Photoshop opens, and now a chat panel opens with it. Adobe started a public beta that drops a bespoke AI Assistant into Photoshop, Premiere, Illustrator, InDesign, and Frame.

In Brief

• OpenAI hires Transformer co-inventor Noam Shazeer ahead of IPO OpenAI recruited Noam Shazeer from Google DeepMind and former Trump AI policy official Dean Ball in the same week. Shazeer co-invented the Transformer architecture underlying modern large language models. The hires land as OpenAI staffs up before a public offering.

• Amazon plans to sell its AI chips to outside data centers AWS is in talks to sell its custom AI chips to other data center operators, moving beyond renting them through its own cloud. CEO Andy Jassy called it a $50 billion opportunity. The plan puts Amazon in more direct competition with Nvidia.

• Z.ai releases GLM-5.2 open weights under MIT license Chinese lab Z.ai published GLM-5.2 on June 16th under an MIT license. The model holds 753B parameters with 40B active via Mixture of Experts, and a 1 million token context window. It is text-only, with vision handled by a separate closed family.

• Baseten raises $1.5B at a $13B valuation for AI inference AI inference startup Baseten is close to finalizing a $1.5 billion round at a $13 billion valuation. The raise comes months after its last mega-round, as demand for inference infrastructure pushes valuations higher.

• General Intuition raises $300M at $2B to train world models General Intuition is in talks for a $300 million round at roughly $2 billion. The startup trains embodied AI and world models on game-clip platform Medal's dataset of 2 billion videos a year from 10 million monthly users.

• OpenAI reasoning model helps diagnose 18 unsolved rare disease cases Researchers used an OpenAI reasoning model to assist physicians diagnosing rare genetic diseases in children. The model identified 18 new diagnoses in cases that had previously gone unsolved.

• Snap spins off its AI video team into a separate company, Dotmo Snap is moving its AI video unit into a new company called Dotmo, citing costs. Dotmo will be staffed by current Snap employees who are leaving to focus on AI video development.

• Snap stock falls after high-priced AR glasses debut Snap unveiled its long-awaited smart glasses, and its stock dropped on the price tag. The hardware reveal did not lift investor confidence in the company's AR strategy.

• OpenAI adds spend controls and usage analytics to ChatGPT Enterprise OpenAI shipped new spend controls and usage analytics for ChatGPT Enterprise customers. The tools let organizations cap and track AI costs across teams.

• Survey finds 60% of US consumers dislike 'AI' in brand messaging A WP VIP survey reports 60% of US consumers see "AI" in brand messaging as a turnoff, and 61% cannot name a brand using AI well. Some 74% say the internet feels less human than a decade ago.

• OpenRouter test pits 11 LLMs in a battle royale, Grok 4.1 Fast wins OpenRouter ran 11 models through 30 games of a 2D battle royale. Grok 4.1 Fast won 43% of matches, while three models won none. The cheapest model beat the most expensive by 27x on cost per win.

• Match survey finds 47% of US singles view AI in dating negatively Match reports about 47% of US singles feel negatively about AI in dating. Many still accept AI for profile edits and conversation starters.

Read the full edition →