Anthropic filed confidentially for an initial public offering with the Securities and Exchange Commission, marking the AI safety company's first step toward going public. The filing comes as Anthropic released Claude Opus 4.8, its latest model, which the company claims delivers 4x improvement in code reliability compared to prior versions. The IPO move reflects growing investor confidence in frontier AI companies, even as debates over their long-term profitability continue.
NVIDIA accelerated its AI infrastructure push at GTC Taipei by unveiling Cosmos 3, bringing its Vera Rubin visual AI system into production, and launching a 1-petaflop AI box targeted at developer laptops. The move extends NVIDIA's reach beyond data-center infrastructure into edge computing and local AI development environments. Vera Rubin represents the company's bet on visual understanding models for industrial applications.
Google retired Gemini 2.0 Flash, consolidating its model lineup. The discontinuation reflects the typical consolidation that follows rapid model proliferation in the LLM market.
On the regulatory front, California's SB 867 passed the state Senate, establishing the first clear ban on AI companion chatbots marketed to children and embedded in toys. The bill targets the emerging category of AI toys and emotionally engaging chatbot companions, which state lawmakers flagged as potential risks to child development. Illinois attempted stricter data-center regulation but saw its proposal stall in committee, signaling regional divergence in AI governance approaches.
The contrast between corporate speed and regulatory pace widened this week. Anthropic's IPO preparation, NVIDIA's infrastructure expansion, and Google's model consolidation all executed within days. Meanwhile, state-level regulation moves glacially. California's SB 867 took months to advance despite broad bipartisan concern about child safety, while Illinois stalled entirely. This dynamic repeats across jurisd