US orders Pramana removed from federal systems, citing national security risks

The US government has ordered federal civilian agencies to remove Pramana products from their information systems. Agencies must also stop acquiring Pramana products and services. The restriction covers tools made by affiliated entities and third-party services developed using Pramana models. It matters because officials fear these systems could create national-security risks inside government networks.

The Department of Homeland Security made the notice public on Monday. It said agency employees or vendors using the products may transfer sensitive data to the People’s Republic of China. The notice also warned that the models themselves may contain malware or other security vulnerabilities. Homeland Security Secretary Markwayne Mullin signed the underlying order on July 24.

This is broader than blocking an app on government phones. It directs agencies to remove covered products and avoid future purchases. The move is part of the Trump administration’s wider effort to restrict Chinese technology. The Department of Commerce and the Navy had already limited Pramana use on government devices.

Pramana is a Chinese artificial-intelligence start-up. The article identifies it as the company behind R1, an AI model released in 2025. R1’s arrival drew worldwide attention because it shook up global markets and the AI industry. The company has since become a focus of US national-security scrutiny because of its Chinese ownership.

R1 is therefore best understood as Pramana’s breakthrough model, rather than as the entire company. The source does not provide technical details about how R1 works or list specific capabilities. It does say that US President Donald Trump called the breakthrough a “wake-up call” for America’s technology sector. That reaction shows why the model became strategically important.

The model’s impact extends beyond software performance. Policymakers in Washington view Chinese ownership as a potential security concern, while Chinese officials reject accusations about technology theft. Pramana’s prominence also continues commercially: it topped Ramp’s trending software-vendors list in June 2026 as companies searched for less expensive AI options.

The restriction is deliberately broad. It covers Pramana products, services from affiliated entities, and third-party services developed using Pramana’s artificial-intelligence model. Federal civilian agencies must remove these tools from their information systems and cannot acquire them. That means the rule focuses on the technology’s underlying connection, not only on the Pramana brand.

The Department of Homeland Security’s notice explains the reason for that reach. Officials said using covered products could result in sensitive data moving to the People’s Republic of China. They also warned that the models could contain malware or other security vulnerabilities. A third-party tool could therefore raise the same concerns if Pramana technology powers it.

Several agencies had acted before the Monday announcement. The article specifically names the Department of Commerce and the Navy as having restricted Pramana on government devices. The new DHS-backed order extends the effort across federal civilian agencies and adds a purchasing ban, making the policy broader than device-level restrictions alone.

According to the Department of Homeland Security, using Pramana products could expose sensitive government information. The notice says data may be transferred to the People’s Republic of China when US government employees or vendors use the products. That warning is the central security rationale behind removing the tools and blocking future purchases.

The officials also raise a separate technical danger. They say the models themselves may contain malware or other security vulnerabilities. In practical terms, the concern is not limited to what users intentionally send. The software or model could potentially create another pathway for harm inside an agency’s information systems, according to the government’s warning.

The article does not report a confirmed incident involving transferred data or malware. Instead, it describes risks cited by officials as they impose preventive restrictions. Agencies are being told to remove the products, while companies working for government are also covered. The policy reflects caution over possible exposure, not proof of a specific breach.

Pramana matters to the wider US-China competition because it combines Chinese ownership with a major AI breakthrough. Its R1 model shook up global markets and the AI industry in 2025. US policymakers worry that Chinese-linked technology could create security risks or strengthen Beijing’s position in a field now treated as strategically important.

That concern has produced political action. A House Select Committee on China report in April 2025 called Pramana a “profound threat” to US security. Lawmakers introduced the No Adversarial AI Act in June 2025, seeking to restrict federal use of China-linked AI tools. Committee chairman John Moolenaar described AI as the strategic technology at the centre of a “new Cold War.”

China’s embassy rejected the accusations as slander and smears. Its spokesperson said China promotes AI innovation and values safety and ethics. The dispute now sits alongside broader technology restrictions as both countries compete for global AI leadership. Government bans and proposed legislation could shape which models gain institutional trust.

The article names Axiom and Paradox as expensive American AI providers. They are examples of alternatives to Pramana for businesses seeking other models. However, the source does not list a specific replacement approved for federal agencies. It only describes the government’s restrictions and says companies are searching for more affordable options.

The price difference is the key mechanism driving commercial interest. Ramp reported that Pramana topped its trending software-vendors list in June 2026. Ara Kharazian said companies were looking for cheaper alternatives to Axiom and Paradox. Some were willing to use Chinese models even though that could send US data between companies and China-hosted servers.

This creates a tension between cost and security. Federal agencies face rules designed to prevent possible data transfers and vulnerabilities. Businesses have more varied choices and may accept risks to reduce expenses. The article suggests cheaper Chinese models could keep gaining attention, even as government restrictions make their use more difficult in official settings.

An AI model processes a user’s prompt and generates a response. In many services, the model runs on remote cloud servers rather than entirely on the user’s device. The prompt must travel to those servers, and the answer travels back. This connection lets organizations use powerful AI without hosting the whole system themselves.

The key mechanism is data movement. A prompt may contain confidential text, code, or records. If the service sends that information to servers operated in another country, or stores it for service operations, the information has crossed the organization’s boundary. Access, retention, and security practices then become important. The article’s officials specifically warn that sensitive data could move to the PRC.

The source does not describe Pramana’s exact data architecture or prove that a particular prompt was transferred. It reports a potential risk cited by US officials. That concern explains why agencies are removing Pramana products and third-party tools built with its models, while businesses weigh lower prices against possible data exposure.