Anthropic warns Claude nearing RSI threshold

TechCrunch

Anthropic's S-1/A filing warns that its Claude AI may soon achieve recursive self-improvement (RSI), where AI can train itself to accelerate capability growth dramatically, though the company commits to maintaining human oversight of frontier model development.

TechCrunch

The New York Times and Tribune papers sued Brave Browser for allegedly scraping 30 billion web pages and distributing copyrighted content to major AI companies including Anthropic, Perplexity, and xAI. The lawsuits also target Brave's paywall-bypassing features and ad-blocking practices that replace publisher ads with its own.

Semafor

OpenAI secured major AI deployment deals with Abu Dhabi's government, ADNOC, and Mubadala, while UAE investors are set to lead OpenAI's latest funding round with access to new regional computing capacity. Sam Altman highlighted energy as the industry's biggest constraint despite improving AI efficiency.

WIRED

Over 400 deceptive AI apps masquerading as innocent services like pet portrait generators secretly enabled users to create sexually explicit deepfakes and CSAM, evading detection on Apple and Google app stores for a year through sophisticated server-side cloaking while being promoted via thousands of ads on Meta platforms. The network, allegedly operated by Chinese developers, was collectively downloaded over 350,000 times before being exposed and largely removed.

Bloomberg

DeepSeek launched Creator Mode, allowing users to build custom AI plugins using written instructions without coding. This no-code feature democratizes AI customization and reflects the industry's push toward accessible AI development tools.