The typo is the least interesting part of this story.
The Summary
- Trump signed an AI safety accord with six tech executives that misspelled "United States" under his signature as "Unites States"
- The voluntary agreement on "Super Intelligence" (Trump's rebrand of AI) is thin on substance, with no binding commitments or enforcement mechanisms
- The accord makes headlines for spelling, but the real story is what wasn't signed: anything with teeth
The Signal
The White House rolled out an AI safety accord Tuesday with the kind of fanfare that suggests something meaningful happened. Six tech executives stood with Trump to sign a document committing to "safely developing" what the administration now calls "Super Intelligence." The press noticed the typo. They should have noticed the empty promises.
Gary Marcus cuts through the theater: this is a voluntary agreement with no enforcement, no metrics, no deadlines, and no consequences for failure. It's the regulatory equivalent of a LinkedIn post about values. Tech companies love these accords because they look like action while requiring none.
"Voluntary commitments from the companies building the most powerful technology of our time are worth exactly what they cost to break: nothing."
The timing matters. We're in the middle of the biggest AI capability jump since GPT-3, with multiple labs racing toward models they openly admit they don't fully understand. Anthropic's Claude, Google's Gemini, and whatever OpenAI is cooking up for GPT-5 are all pushing into territory where the training data, the emergent behaviors, and the safety guarantees are all question marks.
What the accord needed:
- Mandatory safety testing protocols before deployment
- Independent audits of model capabilities and alignment
- Incident reporting requirements when models behave unexpectedly
- Actual liability when things go wrong
What it has: handshakes and a typo.
The "Super Intelligence" rebrand is telling. Trump's language shift from "artificial intelligence" to "Super Intelligence" isn't just marketing. It's framing AI development as a race for dominance rather than a challenge requiring coordination. When you call it Super Intelligence, you're not talking about safety guardrails. You're talking about winning.
The Implication
Voluntary AI safety accords are how governments signal concern without constraining industry. They're designed to fail quietly. Watch what happens six months from now when someone asks which commitments were kept and which weren't. There won't be an answer because there was never a mechanism to check.
If you're building in this space, understand that self-regulation is the current game because actual regulation is stuck in committee. That window won't last forever. The first major AI incident with clear casualties will change the political math overnight. Build like that day is coming, because it is.