AI is ending “pseudonymity” and raising the bar for stablecoin payments on public chains

The last decade taught the internet a hard lesson: “pseudonymous” rarely means “private.” A new shift is accelerating that reality. Large language models (LLMs) are making deanonymization dramatically cheaper and easier to scale.

Recent research shows LLM-driven systems can take a pseudonymous online profile, extract identifying clues from unstructured text, cross-reference public information, and link it back to a real person with high confidence in a meaningful share of cases. The key point is not that deanonymization is new, but that AI turns what used to be manual and expensive into something automated, scalable, and repeatable.

Now apply the same logic to stablecoin payments.

Public blockchains are excellent settlement rails. They provide deterministic execution and reduce counterparty risk compared to closed networks. But they also default to radical transparency. In many designs, settlement may reveal enough metadata to reconstruct sensitive information about users, counterparties, and behaviour.

A concrete example is the use of an on-chain “credit card” product. If the payment rail is on-chain, then a large portion of transaction activity becomes observable in a block explorer. For instance, here is a token transfer view for a single address associated with this activity, showing transfers, timestamps, counterparties, and amounts. If card settlement is routed on-chain, the token transfer log can reflect card-related settlement flows, thus visible by default to anyone.

Even when individual users are not explicitly named, the payment graph is public. And with AI, graphs become profiles. Patterns become predictions. “Just pseudonymous” becomes increasingly linkable and economically exploitable, especially when combined with real-world breadcrumbs that exist in any payments ecosystem (merchants, counterparties, exchange activity, open-source intelligence, and compliance logs).

This is why privacy can no longer be treated as a “nice to have” in stablecoin payments. It becomes a practical requirement. Importantly, privacy is not incompatible with regulation. In regulated finance, the objective is not opacity. It is:

• Confidentiality by default for counterparties, amounts, and payment metadata

• Compliance intact (KYC/KYB, sanctions screening, travel rule messaging, approvals)

• Auditability and controlled disclosure for authorized oversight when required

That is the direction Tessera is built around: enabling bank-grade confidentiality for stablecoin settlement on public chains with zero-counterparty risk, while keeping compliance responsibilities with regulated participants.

As AI continues to compress the cost of inference, institutions will need infrastructure that preserves the benefits of public rails without publishing sensitive payment metadata to the world.