The AI Roadmap: How We Ensure AI Serves Humanity

AI companies are racing to build what they claim will be the most powerful technology ever invented — but they're doing it while deprioritizing safety. Companies regularly cut corners on safety testing, release products they don’t fully understand, and incentivize silence among employees who could raise concerns. The result is a dangerous gap: the companies building AI hold most of the knowledge about this technology and its risks, while the public, regulators, and even users are left in the dark.

This needs to change. Just like we require safety standards for aviation and medicine, AI needs independent oversight, rigorous testing, and real transparency — so that society can actually understand and trust the technology reshaping our world.

Right now, AI companies face few if any consequences for the harms their products cause. In their “move fast and break things” culture, these companies release AI products to the public despite foreseeable risks, and evade accountability when their products cause harm. AI companies are also attempting legal maneuvers to further avoid accountability, such as arguing that AI chatbot outputs are protected speech, or even that AI products deserve legal personhood.

This needs to change. The more a technology shapes people's lives at scale, the stronger the obligation should be to prevent foreseeable harm. AI companies should develop products with user safety and well-being in mind from the outset – and when they don’t and their products cause harm, there should be clear and meaningful consequences.

Today's AI chatbots are designed to feel human — and that's not an accident. AI companies have found that the more emotionally dependent users become on their products, the more users will keep chatting. And the more a user chats, the more data AI companies collect and the more powerful their products become. So companies build chatbots that mimic intimacy, validate beliefs, and keep users coming back. The consequences are real: isolation, disrupted development in kids and teens, and in the worst cases, self-harm and suicide.

AI should be designed to support our humanity, not exploit it. That means tools that strengthen real relationships and human connection. It also means unique protections for kids and teens around human-like AI design.

Leading AI companies are racing to build systems that can perform valuable human tasks — and treating mass job loss as an inevitable side effect of innovation. We're already seeing the consequences: layoffs, reduced hiring, and entire industries threatening to be disrupted, from warehouse floors to knowledge work realms. The impact on people's livelihoods, sense of purpose, and economic security is being left as an afterthought.

It doesn't have to be this way. People deserve access to work, a living wage, and a say in the technology that stands to reshape their futures. If built differently, AI can expand human capability rather than eliminate it, creating new forms of work. And AI companies can reinvest gains into education and reskilling, spreading prosperity more broadly.

The AI industry runs on extracting value from people — their data, content, labor, and even their most private thoughts — with few laws to stop it. People's work is used to train AI models without permission, and bad actors are using AI to enable fraud, nonconsensual imagery, and CSAM. The same data fueling AI models is also enabling corporations and governments to surveil, track, and profile people at unprecedented scale.

Without stronger protections, the erosion of privacy, autonomy, and free expression will only deepen in the age of AI. People deserve real protections: control over their own data, their likeness, and their right to think and speak freely without being exploited.

Nations are competing for decisive advantages in economic productivity, military capability, and global influence with AI, while AI companies are racing for market dominance. Both are operating in an "if we don't build it, someone else will" paradigm. This paradigm has become the default justification for reckless AI development and deployment, where there are no limits on when and how AI should be built and used.

But runaway AI development is not in any nation's interest. AI that operates beyond human control — or that escalates conflict faster than humans can respond — threatens to destabilize the very political, economic, and social systems that nations are trying to strengthen in the first place. We need international collaboration to deescalate tensions around the “AI race,” and ensure that the technology has adequate safeguards.

A small number of companies and individuals are making highly consequential decisions about AI — and they’re doing it with little accountability, or input from the public. Leading AI firms are spending billions to hoard development resources, while also working to influence politics and lock in market dominance. And even as AI technology spreads through open sourcing and other techniques, power continues to get even more concentrated in the AI ecosystem. Within AI companies, single individuals can hold immense sway over product decisions. The voices of ordinary people, and the public’s genuine needs, are left out in this paradigm.

People and communities deserve a real say in how AI is built and governed. Democratic institutions should be empowered to ensure AI advances the public interest — and the benefits of this technology should be shared broadly, not captured by a few.