AI news, 22 September: OpenAI urges US-led global AI standards

OpenAI calls for US leadership on global AI standards. OpenAI on Monday urged the United States to lead an international effort to set technical standards for advanced “frontier” AI systems, including those that could improve themselves. In a blog post timed to the UN General Assembly, the company said shared rules on evaluating models, reporting incidents, and managing risks are essential to avoid a fragmented global system. CEO Sam Altman is due to brief the UN Security Council on Wednesday on AI and international security, according to Reuters and the Financial Times. The move comes amid rising safety concerns and ahead of this week’s Trump-Xi summit.

Gates Foundation launches language-access coalition. The Gates Foundation announced a coalition of about 60 organizations—including Anthropic, Google, the OpenAI Foundation, and others—to expand AI access for people whose languages are underrepresented in current models. The five-year goal is to enable roughly 3.4 billion people to use AI tools in their own languages and voices. Partners will work on open datasets, benchmarks, and applications, building on the foundation’s recent $1 billion AI commitment focused on health, education, and agriculture, according to the foundation and Associated Press reports.

Nvidia CEO rejects AI “doomsday” warnings. Nvidia chief executive Jensen Huang told CBS News there is a “0% chance” AI ends the world by 2030, calling recent extinction warnings “irresponsible” and “not grounded in science.” He argued existing liability and cybersecurity laws are sufficient and new regulation is unnecessary, aligning with a hands-off approach while emphasizing responsible deployment. The comments add to a public split with some industry leaders who have urged a coordinated slowdown.

Amazon blocks Meta’s Muse shopping agent. Amazon has blocked Meta’s new Muse AI agent from accessing and shopping on its retail site after Meta declined a request to remove it. Users saw pop-ups stating the agent violates Amazon’s terms of use. Amazon cited lack of authorization, failure to identify as an agent, and privacy concerns; Meta has said Muse does not see passwords or payment details. The dispute highlights emerging tensions over autonomous AI agents acting on behalf of users.

UN experts caution against apocalyptic rhetoric. Members of a UN independent scientific panel on AI warned against overheated “apocalyptic” language on risks, stressing the need to distinguish known science from speculation. Co-chair Yoshua Bengio and others called for evidence-based discussion. The panel also released a brief noting that traditional safeguards appear to be “unravelling” after incidents involving autonomous agents, urging stronger proactive measures without waiting for full certainty.