Skip to content
AI Primer
breaking

Amp removes proactive ID verification after same-day backlash

Amp reversed its same-day plan to proactively verify user IDs for future frontier-model access and deleted any Stripe verification records. The rollback removes an immediate KYC step, but Amp says governments and model labs could still require identity checks later.

5 min read
Amp removes proactive ID verification after same-day backlash
Amp removes proactive ID verification after same-day backlash

TL;DR

You can read Amp's own security reference, which already emphasizes keeping users on the latest frontier models, and Anthropic's government-directive statement, which says the company had to disable Fable 5 and Mythos 5 for all customers to comply. The oddest detail in the thread is that sqs on Stripe Identity called the integration smooth, then sqs's rollback post killed the whole thing the same day. Amp's identity settings URL now resolves to a sign-in page rather than a live verification explainer.

The rollout lasted one day

Amp announced the feature as a proactive, optional step. The pitch in sqs's launch post was simple: verify now, avoid getting locked out later if a frontier model suddenly ships with identity requirements.

That plan lasted about 16 hours. In sqs's rollback post, sqs said Amp was killing the feature because he did not want to normalize what he called "an iron curtain on the frontier." sqs's deletion note added that users who had already verified were wiped from Stripe as well.

Amp said the pressure was upstream

Amp's core defense was that the restriction was not self-imposed. In sqs's reply on optionality, sqs said verification would only be used if access rules demanded it, and in sqs's terse reply he said Amp had already been asked to implement this once before.

The thread got more specific in pieces:

That last point is the real live wire here. The thread is not just about KYC friction. It is about whether third-party coding agents get the same frontier-model access path as first-party surfaces.

Stripe handled the checks, not Amp

From the start, Amp said it was outsourcing the verification mechanics. sqs's launch post said Stripe would do the identity verification and Amp would see only the outcome, not the underlying documents.

That lines up with Stripe's own Identity docs, which describe a hosted document-verification flow, and with Amp's security reference, which says Stripe handles billing-side personal data for the product. After the rollback, sqs's deletion note said the already-verified records were deleted on Stripe too.

Amp's linked identity settings URL no longer exposes a visible verification flow in public reads. Exa's fetch saw only a sign-in page.

Export controls were only part of the story

One useful clarification from the thread is that Amp did not describe this feature as a direct export-control enforcement tool. In sqs on export controls, sqs said current ID verification would not really work to enforce export controls and called the new rules an emergency measure that would likely be replaced by something more workable.

He still said identity checks would probably be part of whatever comes next. That matters because it separates two claims that often get mashed together:

  • export controls can force abrupt model shutdowns,
  • identity verification can become the practical access gate that survives after the emergency phase.

Anthropic's official statement backs the first half. It says the US government ordered Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, including Anthropic employees who are foreign nationals, and that Anthropic disabled both models globally to comply.

Fable and Mythos were the named precedent

The most concrete new fact in the whole exchange came late in the day. In sqs's reply naming Fable and Mythos, sqs said identity verification was "literally necessary right now" for Fable and Mythos, even if that was only one data point and not the immediate trigger for Amp building the feature.

That gives the story a harder edge than a generic compliance argument. Amp was not preparing for an abstract future. It was reacting to at least one named frontier-model case, and to the possibility that access rules could favor model-lab-owned agents over independent ones, which sqs on lab-owned agents said might miss early access or be cut out entirely.

Further reading

Discussion across the web

Where this story is being discussed, in original context.

On X· 2 threads
TL;DR1 post
Amp said the pressure was upstream3 posts
Share on X