Anthropic and OpenAI sound the alarm on AI safety – and seek to shape how it’s controlled
AI companies Anthropic and OpenAI are advocating for AI safety regulations and independent testing of their systems, while seeking to shape the narrative around their technology's control. Critics suggest these calls for caution serve multiple corporate objectives, including gaining public favor, influencing policy, and securing investments. The article highlights that these discussions often deflect from immediate concerns like data centers' environmental impacts and AI's use in warfare.
The article, an AP report from San Francisco, details how leading artificial intelligence companies like Anthropic and OpenAI are actively advocating for AI safety regulations and independent testing of their advanced systems, while simultaneously attempting to influence the direction of future controls. CEOs from both companies have issued warnings about the potential dangers of powerful AI, suggesting the need for external oversight before deployment, a stance they've articulated in essays, social media, and speeches to the United Nations.
Critics and experts, including former OpenAI geopolitics team lead Sarah Shoker, suggest these efforts serve multiple corporate objectives beyond safety. These include currying public favor, scoring political points ahead of the U.S. midterm elections, positioning for significant stock market listings, and potentially creating barriers for smaller competitors. The narrative push towards "existential risks" is seen by some as a strategic move to divert attention from more immediate, polarizing concerns like the environmental impact of data centers, AI-powered surveillance, and its use in military technology.
The piece highlights that despite calls for regulation, companies like Anthropic and OpenAI are not necessarily pushing for oversight from existing government bodies like the U.S. Center for AI Standards and Innovation, created by President Joe Biden. Instead, they appear to favor establishing their own auditing parameters and selecting specific evaluators. This approach raises questions about independence and credibility among experts like Conrad Stosz, who previously led CAISI. The article also notes that while some, like Nvidia CEO Jensen Huang, dismiss excessive AI alarmism, former OpenAI staff members like Daniel Kokotajlo continue to express grave concerns about unchecked AI advancement.