Startup Subquadratic came out of stealth claiming to have solved a mathematical bottleneck constraining LLMs for nearly a decade. If confirmed, the breakthrough could significantly improve language model efficiency.

ServiceNow published MosaicLeaks, research on privacy vulnerabilities in AI-powered research agents. The study alerts to risks of these agents leaking sensitive information.

The US government forced Anthropic to pull its two latest models over national security concerns after vulnerabilities were discovered in Fable 5. Cybersecurity researchers criticize the decision, arguing similar flaws exist in competing models.

A TechCrunch analysis examines how 30 years of restrictions on cybersecurity software have historically failed. With Anthropic's Mythos model under fire, experts question whether this control strategy is viable in the AI era.

Anthropic faces restrictions distributing Claude Mythos and Fable 5 after running afoul of the Trump administration, but the specific regulations violated remain unclear. The situation highlights the lack of clear frameworks in AI regulation.

The Federal Energy Regulatory Commission (FERC) issued a key decision on large-load interconnection, making it easier for AI data centers and semiconductor facilities to connect to the grid. The measure addresses infrastructure pressure caused by AI.

DeepSeek released V4-Pro and V4-Flash, MoE models supporting 1 million token context with hybrid attention architecture. Both models improve efficiency in long-context scenarios with novel optimizations.

Researchers used an OpenAI reasoning model to diagnose rare genetic diseases in children, identifying 18 new diagnoses in previously unsolved cases. This real-world use case demonstrates the tangible impact of advanced AI in medicine.

Adobe launched custom AI assistants in public beta for Photoshop, Premiere, Illustrator, InDesign, and Frame.io. The initiative is part of Adobe's broader plan to integrate AI across its entire Creative Cloud suite.

OpenAI introduced new usage analytics and spend controls for ChatGPT Enterprise, allowing organizations to manage AI costs more efficiently. This update makes it easier for companies to scale AI adoption with better financial oversight.

OpenAI improved ChatGPT's health and wellness capabilities using GPT-5.5 Instant, with better reasoning and physician-informed evaluations. The changes help the model better understand medical contexts and communicate information more clearly.

Despite the government ban on Fable 5 and Mythos 5, market metrics suggest the restriction has minimal real impact on Anthropic's traction. The episode raises questions about the effectiveness of export controls.

Elastic agreed to acquire DeductiveAI, a startup using AI to catch and resolve software bugs, for up to $85 million. The acquisition reflects the growing market for AI-powered developer experience and code quality tools.

AI inference startup Baseten is close to finalizing a $1.5 billion funding round at a $13 billion valuation. The round reflects fierce competition to position itself in the inference infrastructure boom.

Billionaire Ambani plans to integrate AI into Reliance's calls, apps, and home services serving over 500 million users. This is one of the largest global AI initiatives in telecommunications.

Barret Zoph, OpenAI's head of enterprise AI sales, departed the company after just five months following his return. Zoph had rejoined in January after co-founding Thinking Machines Lab with Mira Murati.

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. A startup claims it broke through a bottleneck that’s holding

One previously unreported SpaceX investor has ties to Chinese military contractors.