Starting July 8, some people who use Claude may be asked to upload a government ID. A passport or a driver's license, a selfie, and the biometric geometry of their face. Anthropic added the requirement to its privacy policy on June 17, and it applies in what the company calls certain circumstances.

The circumstances are narrow, for now. Anthropic told TechCrunch that the checks target a small subset of its tens of millions of monthly users, specifically accounts flagged for potentially fraudulent activity. Rather than ban a flagged account outright, the company offers identity verification as part of an appeals process. It also points to a second driver: state and national age-verification rules that increasingly apply to consumer AI.

A third party, Persona, runs the checks and holds the documents. Anthropic can pull the records for appeals and reviews, and says the data will confirm identity only and will not train its models, according to PYMNTS. Users who fail a first check get more than one attempt.

On its own, this reads like routine trust-and-safety housekeeping. It gets more interesting next to what else Anthropic shipped this week.

Asking for ID is what happens when a model company starts behaving like a platform. The verification gate is the visible edge of a much larger build.

Claude moves into creation

Anthropic released Claude Design, a tool from its Anthropic Labs group that turns a prompt, an image, or a document into designs, prototypes, slides, and marketing material. It applies a team's existing design system so the output stays on brand, converts static mockups into working prototypes, and exports to PowerPoint, PDF, HTML, and Canva. Finished work hands off to Claude Code for engineers to build on. The tool runs on the Opus 4.7 vision model and is in research preview for Pro, Max, Team, and Enterprise subscribers, Anthropic said.

The early numbers are the point. Brilliant said a complex page that took more than 20 prompts in other tools took two in Claude Design. Datadog said work that used to take a week now happens in a single conversation.

Claude is no longer only answering questions about design. It is doing the design. That moves Anthropic up the value chain, into the production work that marketing teams, founders, and product managers pay for. It also widens the surface where a user hands Claude something sensitive: brand assets, unreleased mockups, customer-facing campaigns.

Claude moves into the enterprise

The bigger move is distribution. Anthropic announced a multi-year global alliance with DXC Technology, one of the largest IT services firms in the world, with 115,000 staff across 70 countries and 50 years spent running systems for banks, airlines, insurers, and governments.

DXC will train tens of thousands of Claude-certified engineers and embed them inside customer organizations, putting Claude into the legacy systems that process transactions and claims, the company said. The work centers on four areas: insurance modernization, refactoring old codebases, an always-on security engineer that lives inside a security operations center, and agents embedded in application maintenance.

DXC already runs Claude on its own OASIS platform, which serves more than 50 customers. The company says Claude generated more than 95 percent of OASIS code, reviewed by its engineers, and that development moved roughly 10 times faster. "We are already using Claude across our own operations," DXC chief executive Raul Fernandez said. "Now we are scaling that capability" into customer systems.

Read together, the two announcements describe a company pushing Claude into regulated industries and into the daily creative work of businesses. Both destinations come with rules about who is allowed to do what.

Identity is the connective tissue

Here is why the ID policy and the expansion belong in the same article. As Claude moves into banks, insurers, and brand-sensitive design work, the question of who sits on the other end of the conversation stops being academic. A model that drafts a campaign or touches an insurance claim needs to know its user is real, of age, and authorized.

This is the gap we described in the trust gap in agentic commerce and in the delegation problem: the ability to act is arriving faster than the means to verify who is acting. Anthropic asking a flagged user for a passport is a blunt early version of a control that agentic systems will need in a far more precise form.

When an agent pays a bill or files a claim for you, the counterparty has to trust two things: that you are who you say, and that the agent carries your authority. Mastercard's work on verifiable intent approaches the same problem from the payments side.

Verifying a human with a passport is the easy half. Verifying an agent's authority as it acts across systems it does not own is the half still unsolved.

Anthropic outsourced the human half to Persona and moved on. The agent half does not yet have a Persona to hand it to.

What to watch

Two things. First, scope creep on the ID checks. Anthropic frames verification as an appeals path for flagged accounts, but age-verification law keeps widening, and the same machinery could apply to many more users. Where that line settles will tell us how consumer AI handles identity by default.

Second, who holds the data. Persona stores the documents, not Anthropic. That is a sensible split, and it also adds another party trusted with biometric records. As more AI companies bolt on verification, the identity vendors behind them become quiet infrastructure, holding face scans for millions of people. That is worth watching before it hardens into the default.

The model is no longer the whole product. The platform around it, who can use it, what it can make, and where it can run, is the product now. Identity is how Anthropic draws the lines.

Sources

If an agent acts with your authority, who should have to prove the agent is really yours: you, the platform that runs it, or the network in between?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.