The afternoon briefing.

AI Export Controls. The US government issued an unprecedented export control order against Anthropic, forcing the company to restrict access to its Fable 5 and Mythos 5 models for all foreign nationals, including its own employees. This move, which has no clear legal basis publicly explained, has raised alarms among world leaders at the G7 summit about America's ability to unilaterally cut off access to its AI technology. The incident highlights growing geopolitical tensions and the lack of international consensus on AI governance.

Public Perception of AI. A recent Pew Research poll reveals that while half of Americans now use AI chatbots, up from 33% in 2024, only 16% believe AI will positively impact society. A significant 63% of respondents think AI is advancing too quickly, indicating a widespread public apprehension despite increasing adoption. This sentiment underscores a critical challenge for AI developers and policymakers in building trust and demonstrating beneficial societal impact.

AI Investment and Infrastructure. Major tech players like Amazon, Nvidia, and AMD have collectively invested $310 million in Odyssey ML, a startup focused on building 3D world models, valuing it at $1.45 billion. This substantial backing signals a shift towards next-generation AI beyond traditional language models. Concurrently, an Nvidia-backed optics vendor is quadrupling its wafer output to meet the surging demand for AI interconnects, highlighting the intense pressure on the hardware supply chain.

AI Agents and Automation. Nvidia is leveraging AI coding agents to autonomously direct robot training, enabling self-improvement programs for robotic systems. This advancement in agentic intelligence is further supported by new frameworks like Vercel's open-source Eve, which treats agents as directories, and AWS Context, designed to provide AI agents with a nuanced data lake for improved reasoning. These developments are pushing the boundaries of autonomous systems and intelligent automation.

AI in Cybercrime. Cybercriminals are increasingly exploiting AI to develop sophisticated attacks, as evidenced by a report on Kali365, a platform where hackers use AI to circumvent multi-factor authentication. This comes amidst revelations of alleged Russian-speaking groups compromising tens of thousands of Fortinet firewalls globally through previously known vulnerabilities. Experts warn that advanced AI models with hacking capabilities are becoming the norm, posing significant new threats to digital security.