Demis Hassabis Expects AGI Within 2-3 Years as Anthropic Pitches AI Safety Bill

Demis Hassabis, CEO of Google DeepMind, said he now expects artificial general intelligence (AGI) within 2 to 3 years, accelerating from his earlier 3-to-5-year prediction made in 2026, according to the Wall Street Journal. Hassabis also discussed AI's potential to transform scientific research during the interview.

Anthropic proposed a new AI safety bill that could block release of models capable of killing more than 50 people, though the policy actually prohibits federal agencies from halting a model's release without judicial permission. The proposed framework appears to offer regulatory guardrails while limiting government authority to intervene preemptively against potentially dangerous AI systems.

Meta quietly removed the word "train" from its commercial terms of service following backlash over AI data scraping practices. The company changed its AI scraping policy but notably declined to add language respecting robots.txt protocols, leaving content publishers without clear opt-out mechanisms for Meta's data collection.

US District Judge Zachary Bluestone dismissed a sexual abuse lawsuit filed by Ann Altman against her brother, OpenAI CEO Sam Altman, on Tuesday. According to The Guardian, the judge cited Ann Altman's failure over the past year to appear for an independent medical examination as a key factor in the dismissal of the case alleging childhood abuse.

Hugging Face CEO Clem Delangue warned that the Netflix documentary "AI Gone Rogue," which features him, went "full doomer mode" and is not accurate. Delangue cautioned viewers against sharing the OpenAI-focused hack documentary, which he said misrepresents the current state of AI safety.

Google announced that Gemini's Personal Intelligence features are coming to cars, prompting the company to advise users to audit their Gemini settings ahead of the automotive integration rollout.

Sources: