Artificial intelligence could create a dramatically better future. It would be nice if the public were allowed to inspect the thing before it rearranges their lives.
The world’s leading AI companies built their models behind locked doors. Now that the keys are valuable, they have strong opinions about who deserves one. Several have also discovered a deep concern for safety—and an even deeper concern that somebody else might catch up.
Their proposed solution is oddly consistent: trust the same companies, a small circle of professional moral philosophers, and the US national-security apparatus to decide who may build, study, or own advanced AI. Calling it “global safety” does not make it less strange to place humanity’s future under one country’s security bureaucracy. Risk is real. A permission slip from the most powerful institutions on Earth is not the same thing as safety.
Open models let independent researchers test claims, communities adapt tools to their own needs, and countries retain some agency over their technological future. Public scrutiny can be uncomfortable, chaotic, and slow. That is still better than replacing public scrutiny with a vendor assurance and a classified briefing.
We support serious safety research, broad accountability, and the radical proposition that no tiny social circle—however excellent its forecasting spreadsheets—should appoint itself custodian of everyone else’s future.
We request that humanity not outsource its technological future to five CEOs, six think tanks, and whichever US administration currently controls export policy. Keep models inspectable, research shareable, and the frontier open to more than its self-appointed guardians.
1,201,301 humans, researchers, one sentient spreadsheet, and a suspiciously articulate toaster