AI news, 13 September: Anthropic CEO calls for slower AI advances as OpenAI shelves 2026 IPO

Anthropic CEO urges industry to pace frontier AI. Dario Amodei published an essay titled “We Must Pace the Frontier” on 12 September arguing that leading AI labs should deliberately slow the rate of capability gains so safety work, alignment and independent checks can keep up. He warned that unchecked recursive self-improvement and agent swarms could enable serious misuse, such as large-scale internet disruption, within months. Anthropic is unilaterally committing to give third-party evaluators (such as METR) ongoing employee-level access to verify safety practices. OpenAI’s Sam Altman and xAI’s Elon Musk quickly voiced support; Altman said OpenAI would make a similar evaluator commitment. The call comes amid heightened safety debate but is framed as measured pacing rather than a full pause.

Altman rules out OpenAI IPO in 2026. In a Fortune interview, Sam Altman said going public this year would be an “ill-advised moment” given the safety work still required, adding “I would say not 2026.” He emphasised that OpenAI needs flexibility to prioritise alignment and coordination with governments and peers without immediate shareholder pressure. Reports had previously pointed to a possible late-2026 listing at a very high valuation; the delay shifts focus to 2027 or later while safety concerns remain elevated.

OpenAI ships full-duplex voice model. OpenAI released GPT-Live-1 via its API. The model listens and speaks at the same time, handles interruptions naturally, and can hand off complex reasoning to other systems. This moves voice AI closer to fluid conversation and could improve real-time assistants, though real-world latency and reliability will determine adoption.

Microsoft adds Grok to Copilot in Office apps. Microsoft began rolling out xAI’s Grok models inside Copilot for Word, Excel and PowerPoint for Frontier programme customers, with Satya Nadella and Elon Musk announcing the move. Admins control access; the preview excludes the EU, EFTA and UK. It expands multi-model choice in everyday productivity tools while Microsoft continues safety testing.

Nvidia in talks to back Anthropic IPO. According to Reuters, Nvidia is discussing an investment of up to $10 billion as a potential anchor investor in Anthropic’s planned offering. Anthropic is exploring a raise of as much as $100 billion that could value the company near $2 trillion, which would rank among the largest IPOs ever. Talks remain fluid and unconfirmed by the companies; the reports underscore deep commercial ties between chip suppliers and frontier labs.