An international AI safety agreement is not inevitable. It is built, policy by policy, vote by vote, conversation by conversation. Select your role below to find out what you specifically can do — today, not eventually.
Founders, executives, and operators command credibility that academics and advocates do not. Your choices — what you build, what you refuse, who you partner with — are public signals that move markets and peer behaviour alike.
Make your company's position contractual, board-approved, and verifiable — not aspirational language in a blog post. Commit never to build AGI or ASI systems. A commitment you can be held to is the only kind that counts.
Decline partnerships, joint ventures, and contracts with frontier AI labs racing toward AGI. Publicly explain your reasoning. A business leader who walks away from a deal and says why sends a message that no amount of PR can manufacture.
Beneficial AI is possible without building systems no one can control. If your company works with AI, commit it toward applications with clear, bounded purposes. Make the case by profitable example that safe and useful are not in conflict.
You have access to conferences, publications, and peer networks that most advocates do not. Speak candidly about the stakes from those platforms. Credibility from inside the sector is irreplaceable — it cannot be hired or manufactured from the outside.
Establish clear, written policies about what AI capabilities your organisation will and will not develop or procure. Make them binding on vendors and partners. Normalise the expectation that serious companies have safety standards.
The Nakada Foundation is building the advocacy infrastructure to pass an international AI safety treaty. We are not a research lab. We are a political movement that needs resources to win.
Whether you deploy capital for financial return or philanthropic intent, you hold one of the most direct levers available. Divesting from the race and funding the resistance are both available to you — and both matter.
Remove capital from laboratories explicitly pursuing AGI and ASI. Capital withdrawn is compute not purchased, engineers not hired, timelines not shortened. This is the most direct intervention available to an investor — and it is available today.
Require verifiable safety commitments, compute governance compliance, and independent third-party audits as non-negotiable terms of any AI investment or grant. If they will not agree, that tells you something important about their priorities.
At companies where you hold equity or board seats, put AI safety on the agenda. Push for binding commitments, audit rights, and compliance with emerging international compute governance standards. Boards and LPs respond to large stakeholders. Use that weight.
The AI safety research field receives billions from the labs it studies — a compromised incentive structure. The independent advocacy and policy organisations working for enforceable law receive almost nothing. A single major gift can change the balance of the field.
Investigative journalism into AI lab practices, policy research on treaty frameworks, and cultural work that makes the risk visceral and undeniable — none of this is commercially attractive to produce. Philanthropic capital is the only source for work that is critical but not profitable.
You have access to donors, foundation boards, institutional investors, and policy makers that advocates cannot reach through cold outreach. Convene them. Introduce this cause. An hour of your access is worth a year of effort from an underfunded organisation.
Researchers, engineers, and technical staff at AI labs and technology companies have information, access, and leverage that no external advocate possesses. History shows that workers have changed institutions that seemed impossible to move.
Raise concerns through whatever formal mechanisms exist: safety teams, ombudspersons, ethics review, board access. Put every communication in writing. Create a paper trail before you need one, not after.
You are not alone. Researchers and engineers who are concerned about the direction of frontier AI development exist in every major lab. Find each other. Collective action by technical staff has changed corporate behaviour at AI companies before — and it can again.
People who have disclosed information about unsafe development have done so at significant personal cost. Stand behind them publicly and professionally. Their disclosures make the public case that no outside advocate can make alone — because only insiders know what is actually happening.
In many jurisdictions, whistleblower protections exist for employees who disclose information about activities that put public safety at risk. Understand your rights. Consult a lawyer who specialises in this area. Know your options while you still have time to prepare.
You have skills that are in extraordinary demand. You can get another job — likely a better one. The researchers and engineers building frontier AI systems have more leverage over their employers than almost any workforce in history. Most have not yet used it. The question is whether you will.
If you work at a frontier AI lab and have reason to believe that development is proceeding unsafely — that commitments are not being kept, that risks are being minimised, that the timeline is shorter than publicly disclosed — the question of what you would do is not hypothetical. Think about it now.
No private organisation, however well-funded, can bind sovereign states to limits on AI development. Only governments negotiating with governments can do that. Politicians, civil servants, diplomats, and regulators are the people who will write the law — or fail to.
Require government approval for AI training runs exceeding a defined compute threshold — currently proposed at 10²⁶ floating-point operations. This is technically feasible, independently verifiable, and precedented. A small number of chip manufacturers control the hardware required to train frontier models. License it at that chokepoint.
Formally request that your government table a proposal at the United Nations or a relevant multilateral body to begin negotiations on a binding AI safety treaty. The Montreal Protocol moved from scientific consensus to ratification in under two years. Ambition at this scale is not naive — it is precedented.
AI existential risk is not yet a regular subject of parliamentary or congressional hearings. Put it on the agenda. Invite researchers and former AI lab employees willing to speak candidly. Create the legislative record that future action will require — and that history will consult to understand whether we acted in time.
Push for adequate resources for AI monitoring agencies, independent safety auditors, and the international inspection regime an effective treaty will require. The International Atomic Energy Agency was created within three years of the NPT. Its AI equivalent is achievable on the same timeline if the political will exists.
AI existential risk is not a left-right issue. It is a civilisational issue. Build coalitions across party lines domestically. Reach across to counterparts in other capitals. The treaty will require participation from all major AI-developing nations to be effective — those conversations need to start now, at a political level.
AI companies spend enormous resources on lobbying and access. The policy frameworks they advocate for are designed to preserve commercial freedom of action, not to protect the public. Require independent evidence. Be deeply sceptical of safety standards written by the entities they are meant to constrain.
Artists, journalists, filmmakers, musicians, writers, celebrities, and creators with platforms — you are the people who make abstract threats feel real and urgent. Public opinion is what politicians respond to. You make public opinion.
Statistics and policy papers do not move people. Stories do. Make work — in whatever medium is yours — that makes your audience feel what is at stake: what it would mean to get this wrong, for themselves, for the people they love, for every generation that follows.
Interview AI researchers, former lab employees willing to speak candidly, and policy advocates working on these issues. Your reach can give their message a scale that academic papers and press releases cannot achieve on their own. One interview can do more than a hundred policy briefs.
If you have a platform — a readership, a following, an audience — use it to say plainly that this is the most important issue of your lifetime. Not softened for palatability. Not hedged for commercial reasons. Plainly. Audiences respect directness from people who have something to lose by being direct.
Decline partnerships, endorsements, and sponsored content that benefit frontier AI labs without binding safety commitments. If a lab offers you a deal, turning it down publicly sends a message they cannot buy. They need your audience and your credibility. Use that leverage explicitly.
Science fiction has done more to shape public consciousness about technology than any policy paper in history. The climate movement built its cultural case through art before it built it through law. If you make things — write, direct, compose, illustrate — make the work that makes people reckon with what we are building.
The Nakada Foundation is building a media campaign to shift public opinion on international AI safety. We want to work with creators who understand their platform as a form of political power. Get in touch.
Young people built every major social movement of the twentieth century. This one needs you too — not as bystanders but as organisers, future candidates, and decision-makers whose values and networks are being formed right now.
Start or join a student AI safety organisation. Campaign for your university to divest from frontier AI labs and adopt responsible AI research guidelines. Student pressure has moved universities before — on South Africa, on fossil fuels, on labour practices. It can move them again.
Law, policy, diplomacy, journalism, and government are the fields that will determine whether a treaty gets negotiated and enforced. The AI safety movement desperately needs people in these professions who understand the technical stakes. Your career is one of the most consequential choices you will ever make. Make it count.
Frontier AI labs recruit aggressively from top universities. Declining an offer — and explaining publicly why — carries more weight than any voluntary commitment a lab makes about safety. Your decision to go elsewhere is a form of political speech they cannot counter.
Your classmates are the generation that will be making consequential decisions about AI policy in ten to twenty years. The conversations you have now shape the intuitions they carry later. Do not wait for this to become someone else's problem before you engage with it.
Young people are systematically underrepresented in political contact — and elected officials know it. Write, call, and show up. Tell your representatives that international AI safety is the defining issue of your generation and that you will vote accordingly.
The Nuclear Non-Proliferation Treaty, the Chemical Weapons Convention, the Montreal Protocol — each has lessons for what an AI safety treaty could look like and how it could be verified. Study the history. Become one of the people qualified to design the next one.
Politicians follow voters. Voters who make AI safety a voting issue — who ask the question, show up at hearings, and punish evasion at the ballot box — are the constituency that makes every other action on this page possible.
Ask every candidate where they stand on compute licensing legislation and international AI safety treaty negotiations. If they do not know the issue, inform them. If they dismiss it, remember that at the ballot box. Politicians respond to exactly this kind of pressure — and almost no one is applying it yet.
Phone calls, emails, and in-person constituent visits move elected officials in ways that online petitions do not. Be specific: ask them to support compute licensing legislation and formal international treaty negotiations. Write today — not when it feels more urgent, because by then it may be too late.
Show up in person. Ask your elected officials whether they support an international AI safety treaty. The question itself is an act of political education — for the official, and for everyone else in the room who hears it asked and answered.
Political majorities are built through social networks, not advertisements. Talk to your family, your colleagues, your neighbours. The conversations you have this year shape what politicians feel safe saying next year, and what laws they feel able to pass the year after that.
Donate, volunteer, and canvass for candidates who take AI existential risk seriously. Political campaigns are won by people who show up. This cause needs people willing to show up for candidates who take a position on AI safety before it is the safe position to take.
If you found this meaningful, the most valuable thing you can do right now is expand the constituency for this cause. Send this to one person who has not yet considered these questions. That is how political movements grow: one conversation at a time.
We send briefings when there are developments in international AI policy that require your attention, and when there are specific actions you can take. No noise. Only signal.