What’s New in Artificial Intelligence – September 2025

September 2025 continued to be a pivotal month for AI, with new policy moves, product launches, and shifts in how companies and governments use and regulate intelligent systems.

🔐 Regulation & Policy

• California enacts landmark AI safety law (SB 53)
  On September 29, California passed a law requiring large AI companies to disclose safety protocols, report critical incidents within 15 days, and protect whistleblowers. This marks one of the most aggressive state-level AI regulatory actions in the U.S. 

• Global AI Summit frames new governance agenda
  The Global AI Summit brought together policymakers, researchers, and industry leaders to debate AI’s economic, social, and ethical implications. Topics included inequality, existential risk, workforce disruption, and accelerated regulation.

• UN-led push for AI oversight intensifies
  At the UN General Assembly, leaders and experts stressed the urgency of multilateral AI regulation, launching new initiatives such as a Global Dialogue on AI Governance and calling for clearer “AI red lines.” 

🚀 Products & Industry Moves

• Meta launches “Vibes” AI-powered video feed
  Meta unveiled Vibes, an AI-driven video creation and remix platform integrated with Instagram and Facebook. Users can generate or edit videos by remixing existing content with visuals, music, or text prompts. 

• OpenAI expands Stargate project scope
  OpenAI extended its $500 billion Stargate infrastructure initiative to encompass additional computing projects. The company signaled the establishment of multiple new U.S. data centers in partnership with Oracle, SoftBank, and Nvidia. 

• Zhipu AI offers tempered view on superintelligence
  Zhang Peng, CEO of China’s Zhipu AI, remarked that full artificial superintelligence is unlikely by 2030 and introduced GLM‑4.6, an upgraded model aimed at improved reasoning, coding, and writing performance.

🧬 Advances in AI Applications

• AI predicts flu vaccine strains more accurately than WHO
  A new model named VaxSeer reportedly outperformed WHO’s predictions in retrospective evaluations, using genomic and antigenicity data to recommend vaccine strains with higher accuracy. 

• Salesforce automates support via AI agents
  Salesforce announced that AI agents now handle nearly half of all customer service interactions. The company reduced support staff from 9,000 to 5,000 while claiming improved lead outreach and cost efficiencies.

• New AI tool for heart failure detection in rural patients
  Researchers trained AI models on low-tech ECG data to identify heart failure in rural populations, achieving performance superior to models trained on urban datasets. This holds promise for improving diagnostics in underserved regions.

• UNESCO unveils new AI support programs in Africa
  At the G20 meeting, UNESCO announced initiatives to deploy AI tools in Africa, offering training for civil servants and judicial personnel to bridge capacity gaps in AI adoption. 

📊 Big Picture Trends

• AI boosting global trade by ~40% by 2040
  According to the 2025 World Trade Report, AI implementation could increase trade volume by nearly 40% by 2040, if digital and skills divides are addressed. 

• Latin America launches regionally focused AI model
  A coalition of Latin American nations is rolling out Latam‑GPT, an open-source model tailored to local linguistic and cultural diversity, as part of efforts to democratize AI access. 

• AI discourse on existential risk intensifies
  If Anyone Builds It, Everyone Dies, a new book by Eliezer Yudkowsky and Nate Soares, examines the existential dangers of superintelligence and has become a focal point in discussions about AI futures. 

✅ What September Tells Us

• AI regulation is moving from theory to enforcement — states and nations are now imposing binding oversight.

• AI-driven content creation and remixing are entering the mainstream.

• AI models are maturing in domains such as healthcare, public services, and enterprise automation.

• A tension grows between ambitious claims of superintelligence and more realistic forecasting.

• Regional AI initiatives (Latin America, Africa) are gaining momentum — pushing beyond just U.S./China dominance.