Perplexity launches secure AI for the U.S. government

Perplexity announced the launch of Perplexity for Government on September 8, 2025, a program designed to make it easier for U.S. federal agencies to use AI with automatic data protections and access to advanced models. The company says this initiative centralizes its support for public administration and adapts its products to the needs of the public sector. (perplexity.ai)

What Perplexity announced

The news has two main pieces:

• Secure-by-default access: Perplexity says it will recognize requests coming from federal agencies and automatically apply zero data usage policies — meaning the data won’t be retained to train models. This applies both to authenticated users and to visits without a session. (perplexity.ai)

• A tailored enterprise edition: Perplexity Enterprise Pro for Government, a customized edition of its enterprise platform that would combine Perplexity’s own models and third-party models, offered to agencies via government procurement vehicles, with a nominal price announced for the first 15 months. (perplexity.ai)

Sounds technical? Think of an analyst in a federal office who asks an AI tool to summarize legislation or prepare a briefing. With this announcement, Perplexity promises those queries won’t be used to improve public models or stored for other purposes. (perplexity.ai)

Why it matters (and why now)

This comes at a moment when U.S. public policy is pushing for both adoption and governance of AI. The White House recently published the "America's AI Action Plan" and management memoranda that push to accelerate responsible AI use across the public sector. That pressure pushes companies and agencies to define solutions that pair leading capabilities with security and compliance controls. (whitehouse.gov)

In other words: it’s not just about putting a chatbot on an intranet. It’s about balancing usefulness and risk when information can be sensitive or classified. Perplexity is betting on an experience that’s "secure-by-default" for federal employees. (perplexity.ai)

How it promises to protect data

Perplexity describes a technical and operational approach: identify traffic coming from agencies via network ranges and other methods, apply zero data usage, and automatically route those requests to its best models to give more powerful answers without retaining the interaction. Agencies can also provide additional network ranges to widen coverage. (perplexity.ai)

Perplexity says these protections apply whether the user is authenticated or not. (perplexity.ai)

This doesn’t remove the need for each agency’s own evaluation: internal policies, data classification, and procurement agreements remain critical. But it lowers a technical barrier — the fear that internal queries could leak into public model training. (perplexity.ai, whitehouse.gov)

Perplexity Enterprise Pro for Government: what it offers and how much it will cost

The company also announced a tailored enterprise offering, Perplexity Enterprise Pro for Government. According to the post, the plan is to offer the platform via federal purchasing vehicles like the GSA Multiple Award Schedule, and it mentions a nominal cost of USD 0.25 per agency for the first 15 months to encourage broad adoption. (perplexity.ai)

The idea is to combine Perplexity’s models and third-party models to provide options and reduce the risk of relying on a single provider or model. For many agencies this can be attractive: more technical flexibility and the ability to integrate institutional knowledge securely. (perplexity.ai)

What this means for government employees, vendors, and the rest of us

For federal employees: more access to AI tools that promise privacy by default, but you’ll still need internal governance and training to use AI responsibly. (perplexity.ai, whitehouse.gov)

For contractors and vendors: there’s a window of opportunity to integrate services that comply with these guarantees and participate in government contracts. If you work at a company that supplies software to the public sector, this changes the compliance requirements to consider. (perplexity.ai)

For the general public: the promise is that administrations can use AI to improve services without sacrificing the security of sensitive information. What will matter in practice are clear policies, audits, and transparency. (whitehouse.gov, perplexity.ai)

Reasonable doubts and points to watch

• Traffic identification: relying on network ranges can cover much of the usage, but it’s not a total guarantee. You’ll want to see how this works in practice. (perplexity.ai)

• External audit and verification: will there be independent mechanisms to confirm that zero data usage is actually enforced? This is where oversight and auditing matter. (whitehouse.gov, perplexity.ai)

• Scope of the service: the announcement promises broad coverage, but effective adoption will require agreements, internal guidance, and likely training. (perplexity.ai)

How to get more information

Perplexity enabled a government-dedicated hub and a contact channel in its official post so interested agencies can consult details and provide additional network ranges. If you work in an agency or at a company that contracts with government, that’s the first point of contact the company announced. (perplexity.ai)

Final reflection

Is this a before-and-after moment for AI in the public sector? It’s an important step: it promises cutting-edge capabilities with default protections, and it arrives aligned with the federal push on AI. But the promise will be tested in implementation: identifying traffic, auditing compliance, and ensuring the platform fits each agency’s rules and needs.

The announcement raises practical questions and real opportunities. If you’re a public servant, a contractor, or just curious about how AI affects public administration, it’s worth following this program’s evolution and reviewing the agreements and controls that are defined in the coming months. (perplexity.ai, whitehouse.gov)