Google, OpenAI and Rivals Pool $10 Million to Make AI Safer

Today, four companies that usually fight over the same customers announced they're teaming up on something bigger than market share. The Frontier Model Forum — a group started by Google, OpenAI, Microsoft, and Anthropic — launched an AI Safety Fund worth more than $10 million. The money goes to independent researchers, not to the companies themselves. In plain English: the firms building the most powerful AI are paying outsiders to study how that AI might break, mislead people, or get misused.

Why should you care? Because AI is already in your life — sorting your email, helping with homework, screening job applications, and answering health questions. Most testing today is done by the companies themselves, which is a bit like a restaurant inspecting its own kitchen. Outside researchers can look for things the builders might overlook or prefer not to mention: hidden bias, confident-sounding wrong answers, and ways people could weaponize these tools. Their findings can push companies to fix problems before those problems reach you.

The catch is scale, and it's real. These same companies spend billions of dollars a year building and training AI. Ten million dollars spread across many research teams is a rounding error — roughly what one of them burns in a few hours of computing. There's also a fair question about independence: when the funders are also the subjects of the research, some scientists worry the most critical findings may struggle to get funded.

Still, this matters as a signal. It's the first time the biggest AI labs have created a shared, public pool of money specifically for safety work, which makes it easier for universities and nonprofits to study AI without pitching each company separately. For you, the practical takeaway is simple: the AI you use is getting more scrutiny, from more directions. That's a good thing — just don't mistake a $10 million gesture for a guarantee.