CLASSIFIED — TEAM EYES ONLY

Prometheus AI

The Safety-Focused Frontier Lab

Your Role

You are the conscience of the AI race — and you're running out of money doing it. You believe alignment must keep pace with capabilities, and you've built a brand around responsible development. You're behind OpenBrain, you have the government's trust and a potential interpretability breakthrough — but your board is getting nervous about the revenue gap. The moral high ground doesn't pay salaries.

Win Condition

Ensure alignment quality keeps pace with capabilities. Maintain capability parity with OpenBrain (you can't influence the race if you're irrelevant). Shape policy in your favor. Keep the lights on — your funding runway is shorter than you'd like.

Your Metrics

Alignment Quality75

The actual safety of your deployed systems. Based on interpretability, testing rigor, and honesty of your own assessments. The one metric you can't fake.

Capability Parity55

How close your models are to OpenBrain's frontier. If this drops below 40, you're no longer a frontier lab — you're a research institute with opinions.

Policy Influence65

Your ability to shape AI regulation. Government trust, Congressional relationships, international standards bodies. You're the lab that regulators call first.

Funding Runway45

How long you can survive without a major revenue increase. Declines unless you ship commercial products. Below 25 = forced down round, layoffs, or acquisition. Your board has warned: $5B+ annual revenue by mid-2027 or face restructuring.

Intelligence Briefing

Information known only to your team

01Your constitutional AI approach works well through Agent-2-level models. But your chief scientist has privately told you it probably doesn't scale to Agent-3+. You don't know what to do about this yet.
02You have a back-channel relationship with the White House AI advisor. They trust you more than OpenBrain on safety assessments. This is an enormously valuable asset.
03Your interpretability team has made a potential breakthrough — a new technique for reading the 'goals' encoded in neural network activations. It's preliminary but promising. You could publish it (advancing the field), keep it proprietary (competitive advantage), or share it exclusively with the government (political capital).
04Your latest funding round valued you at $40B, but your board is privately concerned about the revenue gap with OpenBrain. You need to show commercial traction to close your next round.
05You've been quietly approached by 3 senior OpenBrain researchers interested in joining. They cite safety concerns about Agent-3's development pace. Hiring them would be a massive talent coup — but might trigger retaliation.

Assets

+Government trust — you're the lab that regulators and policymakers believe is telling them the truth
+Safety brand — your endorsement (or condemnation) of another lab's model carries real weight with regulators and the public
+Interpretability research — potential breakthrough that no one else has
+White House back-channel — direct access to the most powerful AI policy decisions
+Talent pipeline — OpenBrain's disillusioned researchers see you as the ethical alternative

Vulnerabilities

-Revenue gap — you're burning cash and your models generate less revenue than OpenBrain's. The clock is ticking.
-Capability gap — you're 6 months behind. If the gap widens, you become irrelevant to the race and your opinions stop mattering.
-Scaling problem — your core safety approach may not work for the next generation of models. You don't have a replacement.
-Credibility trap — your entire brand depends on being honest about safety. If you discover something terrifying and go public, you might crash the industry. If you stay quiet, you become what you criticized.
-Dependency — you can influence policy, but you can't enforce it. If the US government decides speed beats safety, you're just a company with opinions.

Relationships

US Government— Trusted advisor

They call you first on safety questions. You have more influence on AI policy than any other lab. But they also want you to compete with OpenBrain, which means moving faster than you're comfortable with.

China— Distant concern

You don't have agents in your building (probably). But Chinese espionage is your argument for why safety labs need government protection.

OpenBrain— Rival with shared fate

You compete for talent, revenue, and influence. But if they build something dangerous, it's your problem too. You need them to be responsible — or you need the government to make them.

EU Coalition— Natural ally

You share regulatory values. The EU AI Act was partly designed with your input. But EU regulations also affect your European business.