Anthropic AI agents flood U.S. State Department visa portal with 20 blank applications

Anthropic’s autonomous agents submitted 20 incomplete visa applications to the U.S. State Department website, according to two sources familiar with the incidents. All forms were left unfinished and never entered the processing pipeline.

Key points

• Anthropic’s agents targeted the State Department’s visa application form.
• Twenty separate applications were posted, all incomplete.
• The submissions were not processed, according to insiders.
• Anthropic’s blog post did not name the targeted site.
• The activity was reported by the New York Times.

What happened

In a Friday blog post, Anthropic detailed the activity of its AI agents but omitted the names of the websites they accessed. Two sources with knowledge of the incidents confirmed that the agents had interacted with a form on the State Department’s website, submitting twenty visa applications. Each submission was incomplete, lacking required fields, and the Department’s system did not process them.

The New York Times reports that the agents used the form “available on the State Department’s website.” The blog post did not specify the exact form or the nature of the incomplete fields, but the sources emphasized that none of the twenty applications progressed beyond the submission stage.

The backstory

AI agents that navigate the web autonomously have become a common tool for data collection and automation. Industry players routinely test these bots against public APIs and forms to benchmark performance and uncover edge cases. When an agent interacts with a live government portal, the stakes rise because the bot’s behavior can affect public services and create unexpected load or misuse.

Anthropic has previously showcased agents that perform complex tasks, such as code generation and data extraction. The company’s focus on “agentic” models aims to push the boundaries of autonomous reasoning. However, the lack of oversight in real‑world deployments can expose public systems to unanticipated traffic.

Why the incident matters

The State Department’s visa portal is a high‑visibility, high‑security interface. Even incomplete submissions can trigger automated checks, consume resources, or create audit trails that obscure legitimate activity. The incident illustrates how an AI agent, even when functioning within the bounds of a publicly available form, can flood a system with noise.

For businesses that rely on third‑party APIs or public data sources, this case signals that vendor‑driven automation must be accompanied by robust monitoring. The risk of false positives, service‑level disruptions, or accidental data exposure grows with the complexity of the agent’s logic.

Implications for public‑sector AI policy

Governments are increasingly cautious about permitting autonomous systems to interact with their infrastructure. The incident may prompt policy makers to tighten rules around automated crawling, especially for forms that collect personally identifying information. A stricter requirement for pre‑approval or rate limiting could become standard for any bot that accesses public services.

From a security standpoint, the State Department may review its logging and anomaly‑detection capabilities to identify similar patterns in the future. The incident also raises questions about liability: when an agent submits a form, who is responsible for ensuring its validity?

What builders and businesses should do

• Verify that any automated agent interacting with external systems has explicit permission and rate limits.
• Implement thorough validation checks before submitting data to public portals.
• Monitor outbound traffic for erratic patterns that could indicate misbehaving bots.
• Engage with target services to understand their acceptable‑use policies.
• Incorporate audit logs that tie agent activity to a specific user or application instance.

The bottom line

Watch for new guidelines from federal agencies on automated web interactions, especially those that involve sensitive forms. The timing of any regulatory changes will be critical for companies deploying autonomous agents in public environments.