Charlie Munger famously said, “show me the incentives and I’ll show you the outcome.” Individuals and organizations alike are almost always driven to optimize for the specific incentives they’re offered. That sentiment has never been more true than it is in AI.
This week at the Paris AI Action Summit, Vice President JD Vance made a clear point about the incentives of AI incumbents:
“When a massive incumbent comes to us, asking us for safety regulations, we ought to ask whether that safety regulation is for the benefit of our people. or whether it's for the benefit of the incumbent.”
The Vice President’s point bears that unique quality of something that feels obvious, but often escapes the common zeitgeist. AI safety has become the default rallying cry for, what feels like, the attempted regulatory capture of the largest and most well-funded AI companies. Now, why would they do that? The colloquial phrase is “pulling up the ladder behind them.” Large AI platforms are incentivized to make sure its harder for other companies to compete with them.
The same is true of the large tech incumbents that are pouring billions into the AI race. In fact, you can trace a line directly between each large incumbents stance on AI safety and their financial incentives.
Who’s in favor of proprietary, controlled ecosystems? Amazon, Google, Microsoft. Each have invested billions into closed, proprietary platforms like Anthropic and OpenAI. Why? Because the more control they have, the more they can ensure they sell more compute as part of AWS, GCP, or Azure.
Who’s in favor of open source models? Meta. The company is set to invest $60 billion in infrastructure to support Llama, completely open-source and available to everyone. Why? Because Meta is primarily focused on ensuring its products get better — Instagram, WhatsApp, Facebook.
Increasingly, companies from foundation model platforms like OpenAI and Anthropic to hyperscalers, like Amazon and Microsoft, are having more and more light shed on their incentives. Most recently, that light came in the form of DeepSeek. When the news of DeepSeek caught fire and wiped out hundreds of billions of dollars in market cap from companies like Nvidia, it rattled an established paradigm within OpenAI in particular.
In an Ask Me Anything shortly after DeepSeek’s big day, Sam Altman shared an honest reaction to the pressure the company was feeling. Regardless of the suspicious details around the creation of DeepSeek’s models, the reality of how much value it provided in being open was the light bringing OpenAI’s incentives into focus.
This feels like a shocking admission from the person who, to date, has been the standard bearer for closed, proprietary models. This brings us back to Vice President Vance’s comments. We’re starting to see much of the AI safety bluster for what it is; an attempt at building a protective shield around the competitive moats of large incumbents. When consumer expectations shift and public pressure mounts, its not the value system that stands the pressure — it’s the incentive.
