AI Newsletter

Nvidia secures a dominant position in AI infrastructure through a massive guarantee ($105 billion) and exclusive chip supply contracts, while OpenAI builds massive computing capacity. This significantly strengthens Nvidia's monopoly position in AI chips and signals that tech companies are investing trillions in AI hardware – a sign of both growth potential and possible overcapacity.

Anthropic's watermarking technology for Claude is intended to mark AI-generated content but could affect text quality and create new legal transparency issues. The criticism jeopardizes trust in the technology and could increase regulatory pressure, weakening Anthropic's competitive position against OpenAI and other LLM providers.

Leading mathematicians confirm LLMs as efficient computational tools for known methods but deny them creative breakthroughs. This limits AI's applicability in highly specialized R&D areas where true innovation provides competitive advantage.

OpenAI has dissolved its specialized safety team for catastrophic AI risks and distributed its tasks to existing departments, resulting in the departure of several safety experts. This signals a deprioritization of risk assessments and could damage the company's credibility with regulators and investors. Internal resistance suggests tensions between safety claims and speed orientation.

Anthropic revealed a critical security lapse: a filter against misuse for biological and chemical weapons was inactive for ~12 months, while 133 million user requests were processed unprotected. This not only jeopardizes corporate trust and product reputation but also significantly increases regulatory and legal risks.