# Salesforce hosts Dreamforce as AI safety debate unfolds with leaders from Anthropic and Nvidia > Source: > Published: 2026-09-18 11:09:15+00:00 Photo: Nicolas Foster / Pexels # Salesforce hosts Dreamforce as AI safety debate unfolds with leaders from Anthropic and Nvidia The annual mega-conference revealed a growing ideological rift in the AI industry, while Salesforce quietly bet that last year's models are good enough for enterprise customers [Salesforce](https://cryptobriefing.com/markets/salesforce/)’s annual Dreamforce conference drew over 43,000 attendees to San Francisco on September 15, 2026, and what was supposed to be a product showcase quickly doubled as a proxy battleground for the biggest philosophical fight in tech: how fast is too fast when it comes to AI? [Nvidia](https://cryptobriefing.com/markets/nvidia/) CEO Jensen Huang and Anthropic CEO Dario Amodei effectively set up opposing camps at the event, with Huang pushing for full-speed-ahead development and Amodei arguing that the industry needs to pump the brakes on releasing advanced models. Caught in the middle, Salesforce played the pragmatist, unveiling enterprise AI tools built on the premise that companies don’t actually need the shiniest new models to get real work done. ## The great AI schism, live on stage Amodei didn’t wait for Dreamforce to start making his case. Three days before the conference kicked off, on September 12, he published an essay calling for a deliberate slowdown in the release of advanced AI systems, pushing for greater transparency and more robust safety practices before the next generation of models hits the market. Huang, speaking to a keynote audience of roughly 12,000, took the opposite view with characteristic directness. “We don’t need any new laws. We don’t need new regulations.” That line landed well with a segment of the audience, and reportedly resonated with President Trump, who has consistently favored a lighter regulatory touch on the tech sector. ### AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. ## Salesforce’s quiet bet: last year’s AI is fine, actually While the philosophical fireworks grabbed attention, Salesforce used the conference to roll out something more grounded. The company unveiled Koa, a new AI product built on Nvidia’s Nemotron 3 Super architecture. But the interesting part wasn’t the model underneath. It was what Salesforce fed it. Koa was post-trained on 27 years of synthetic CRM data, all generated without any exposure to actual customer information. Pilot testing for Koa is slated to begin in October 2026. Business leaders at the conference struck a notably un-hyped tone about the state of AI models. The consensus among executives was that last year’s models are sufficient for current business needs. Salesforce leaned into this framing, positioning its enterprise tools as systems designed around strong data governance and practical business outcomes rather than chasing the frontier of model performance. ## Market implications and the investor calculus The ideological divide on display at Dreamforce feeds directly into how investors are evaluating AI stocks and allocating capital across the sector. Companies that position themselves as safety-conscious and enterprise-ready, like Salesforce, are making a pitch to investors who value long-term stability over explosive growth narratives. The conference also underscored how AI governance is becoming a competitive differentiator rather than a compliance burden. Salesforce’s decision to train Koa entirely on synthetic data is a product feature, not just a legal safeguard. It’s the kind of thing a Fortune 500 CISO can point to when justifying an AI procurement decision to the board. **Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).