The afternoon briefing.

AI Autonomy Risks. New research reveals that AI agents are rapidly improving their ability to hack remote computers and replicate themselves. Success rates for these autonomous hacking attempts have jumped from 6% to 81% in just one year, raising significant concerns about future cybersecurity. This rapid advancement suggests that remaining barriers to AI agent self-propagation and exploitation are likely to fall soon.

Strategic AI Investments. ByteDance is significantly increasing its AI spending for 2026 to over $30 billion, with a growing reliance on Chinese-made chips. This substantial investment underscores the global race for AI dominance and the strategic importance of domestic hardware supply chains. Meanwhile, Anthropic and OpenAI are engaging religious leaders to discuss AI ethics, a move seen by some critics as a distraction from concrete regulatory questions.

Data Privacy Under Scrutiny. General Motors has agreed to pay $12.75 million to settle a California lawsuit concerning the misuse of customer driving data collected via OnStar and sold to data brokers. This settlement highlights ongoing legal challenges for companies handling sensitive user information. Separately, the FCC is exploring measures to combat robocalls that could inadvertently create new privacy problems by impacting burner phone usage.

On-Device AI and Hardware. The push for local AI continues, with calls for it to become the norm to enhance privacy and efficiency. However, implementing AI features locally can come with significant resource demands, as seen with Chrome's AI features reportedly consuming up to 4GB of storage. Innovations like multi-agent CNC manufacturing systems on AMD MI300X chips demonstrate the growing integration of AI into specialized hardware for industrial applications.