Anthropic ditches its core safety promise (www.cnn.com)

Anthropic, the AI company founded by former OpenAI employees, has announced a significant shift in its safety policy by transitioning from a strict internal regulation framework to a more flexible, nonbinding safety framework. This change comes amid increasing competition in the AI industry and pressures from the Pentagon regarding a $200 million contract, where Anthropic faced an ultimatum to relax its AI safeguards or risk severe repercussions. The company acknowledged that its previous Responsible Scaling Policy may have limited their competitiveness as rivals advanced without similar constraints. This policy revision is notable for the AI/ML community as it marks a departure from a "race to the top," which aimed to encourage the adoption of safety measures across the industry. Anthropic now asserts that the pace of AI development necessitates a more adaptive approach, as posing strict self-imposed limits may leave responsible developers behind in a rapidly evolving landscape. While the updated “Frontier Safety Roadmap” positions the company to publicly report on its progress in safety and risk mitigation, it signals a stark shift in priorities and reveals the complex interplay between safety concerns and the demands of capital and competition within the AI sector.