Can a US developer still call the DeepSeek API? Can a state employee in Texas open the DeepSeek app on a personal phone? The short answer to the first question is yes; the second is more complicated. The **DeepSeek US restrictions** landscape that took shape in early 2025 — after the R1 release triggered what Marc Andreessen called an “AI Sputnik moment” — has hardened into a patchwork of state executive orders, agency directives, and pending federal bills, rather than a single national ban. As of April 24, 2026, no law stops a private citizen or company from downloading the app, running the open-weight models, or paying for API access. What’s restricted is narrower, and the distinctions matter. This article maps exactly who can and can’t use DeepSeek in the United States today.

The bottom line: no general US ban on DeepSeek

There is no federal law prohibiting Americans from using DeepSeek. The restrictions that exist target government devices, government networks, and federal contractors — not private individuals, businesses, or open-source developers building on the DeepSeek V4 weights. If you are a US-based developer, you can still hit the POST /chat/completions endpoint at https://api.deepseek.com, and you can still download the MIT-licensed V4 weights from Hugging Face. What has changed since the R1 launch in January 2025 is the compliance posture expected of public-sector and regulated workloads.

This article summarises the actions taken by US states, federal agencies, and Congress, separates rumour from codified restriction, and explains what the rules mean in practice for developers, enterprises, and researchers.

Timeline: how US restrictions emerged

The regulatory reaction moved fast after R1 landed. DeepSeek became the most downloaded app in the United States in late January 2025, and its Chinese ownership quickly raised data security concerns among US federal and state officials. Within weeks, three tracks opened in parallel: state executive orders, federal agency internal bans, and congressional legislation.

• January 31, 2025: Texas became the first state to act, with Governor Greg Abbott announcing that the state would not allow the use of AI and social media apps affiliated with the People’s Republic of China on government-issued devices — the banned list included DeepSeek, RedNote, Webull, Tiger Brokers, Moomoo, and Lemon8.
• February 6, 2025: US Representatives Josh Gottheimer (NJ-5) and Darin LaHood (IL-16) introduced the bipartisan “No DeepSeek on Government Devices Act,” prohibiting federal employees from using DeepSeek on government-issued devices.
• February 10, 2025: New York Governor Kathy Hochul announced a statewide ban on the use of DeepSeek’s AI application on state government-managed devices and networks, citing serious concerns about foreign government surveillance and censorship.
• February 11, 2025: Virginia became the third state to prohibit DeepSeek on government devices with Governor Glenn Youngkin’s signing of Executive Order 46, which prohibits state employees and contractors from downloading or using DeepSeek on state-owned equipment or over state wireless networks.
• Late February 2025: A bipartisan trio of senators — Jacky Rosen, Jon Husted, and Pete Ricketts — introduced a Senate companion bill to ban DeepSeek from all federal government devices and networks, following the House bill earlier that month.
• March 2025: Alabama and Oklahoma joined the list, blocking AI tools with overseas ties on government devices, citing a lack of security and data-collection practices.

State-by-state snapshot

State actions have been the most concrete layer of the DeepSeek US restrictions regime. The table below covers the executive orders and memoranda that are publicly documented. Legislative bills in other states (for example, Georgia’s SB 104 and Kansas’s HB 2313) remain in the pipeline. Coverage is narrower than it looks — these rules bind state employees and contractors, not the general public.

A coalition of 21 state attorneys general has also called on Congress to enact comprehensive legislation addressing DeepSeek. State-level university systems have followed their governors — the University at Buffalo aligns with a statewide and SUNY ban on DeepSeek AI, meaning faculty, staff, and students must not download or access DeepSeek on UB-owned devices or over UB networks — which extends the practical footprint to public research environments.

What state bans actually prohibit

Read the Virginia order carefully and the scope becomes clear. It directs that no employee or contract employee of any agency of the Commonwealth shall download or use DeepSeek on government-issued devices, including state cell phones, laptops, or other devices capable of connecting to the internet, and prohibits downloading or accessing the DeepSeek app on Commonwealth networks. What it does not do is restrict private residents from using DeepSeek on their own devices and home Wi-Fi.

Federal actions and the legislative track

At the federal level, two things are true at once: individual agencies have already acted, and Congress has not yet passed a single comprehensive law.

Agency-level bans

Several federal bodies moved on their own authority through 2025. Several US Commerce Department bureaus prohibited the use of DeepSeek on government-issued devices, with a mass email instructing employees to refrain from downloading, viewing, or accessing any applications, desktop apps, or websites associated with DeepSeek. The US Navy also moved to restrict the application. DeepSeek has also been barred from use on Senate and House devices, according to Axios.

Pending legislation

Three bills matter most for readers tracking the DeepSeek US restrictions picture:

1. No DeepSeek on Government Devices Act (HR 1121 / S 765). The bill directs the Office of Management and Budget to develop standards and guidance requiring removal of the DeepSeek application from federal agency information technology, applying also to any successor application developed by High-Flyer or entities it owns. It carries limited exceptions for law enforcement, national security, and security research, with risk-mitigation documentation required when those exceptions are invoked. As of the latest public tracking data, both the House and Senate versions remain in committee.
2. Decoupling America’s Artificial Intelligence Capabilities from China Act (Sen. Hawley). This separate bill seeks to penalise the importation of technology or intellectual property developed in China, with penalties including up to 20 years in prison and fines of up to $100 million for organisations that violate it; it does not explicitly mention DeepSeek but would, if enacted, have a much broader effect than the device-level bans.
3. Protection Against Foreign Adversarial Artificial Intelligence Act of 2025 (Sens. Cassidy and Rosen). This bill would prohibit federal contractors with an active federal contract from using DeepSeek — and any successor application developed by High-Flyer — for the fulfillment, assistance, execution, or support of a contract with a US federal agency.

Anyone building production systems touching federal contracts should read these carefully, because any future DeepSeek prohibition could be extended by the FAR Council to all federal contractors and may not exempt commercial-item contracts under FAR Part 12 or contracts below the simplified acquisition threshold, leaving contractors with questions about implementation scope.

Why lawmakers are worried: the stated concerns

The national-security argument is consistent across orders, memos, and bill statements. It has three main threads.

Data handling under Chinese law

Security experts have raised privacy and security concerns about DeepSeek, noting that all data it collects is stored on servers in China, potentially exposing it to surveillance by the Chinese government under that country’s data laws. This is not a speculative point: Chinese law requires companies to cooperate with national intelligence work on legal request.

The China Mobile code link

The Associated Press reported that DeepSeek has computer code that could send some user login information to China Mobile, a Chinese state-owned telecommunications company that has been barred from operating in the United States, according to the security research firm Feroot. This finding appears repeatedly in congressional statements and state press releases.

Security incidents at the company

A security issue at DeepSeek has exposed sensitive internal data, researchers at Wiz found. The incident, widely reported in early 2025, compounded concerns about how DeepSeek handles user data.

These points shape what caution looks like for US teams: the concerns are real enough to justify keeping the hosted API out of anything touching government contracts, regulated data, or health records, while not severe enough to warrant a consumer-level ban. For context on how to reason about these trade-offs, our take on DeepSeek privacy and the broader question of whether DeepSeek is safe goes deeper.

What developers can still do in the US

If you are a US-based individual or private company not subject to state or federal device rules, nothing in current law stops you from using DeepSeek. Three access paths remain open:

• Hosted API. The DeepSeek platform serves deepseek-v4-pro and deepseek-v4-flash from its own infrastructure. Chat requests hit POST /chat/completions, the OpenAI-compatible endpoint at base_url="https://api.deepseek.com". Legacy IDs (deepseek-chat, deepseek-reasoner) still route to V4-Flash until 2026-07-24 15:59 UTC, after which they fail. See the DeepSeek API documentation for the full reference.
• Open-weight self-hosting. Both V4 tiers ship open weights under the MIT license. A US company with its own GPUs (or renting on a US-hosted cloud) can serve the model without sending any inference data to DeepSeek’s servers. The install DeepSeek locally tutorial walks through that path.
• Third-party US hosting. Several US-based inference providers mirror DeepSeek weights on domestic infrastructure. Data sovereignty concerns largely collapse in that model, though licensing and benchmark parity should be verified per provider.

A minimal Python call using the OpenAI SDK, which matches DeepSeek’s wire format:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.deepseek.com",
    api_key="YOUR_DEEPSEEK_KEY",
)

resp = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Summarise the US state bans."}],
    temperature=1.3,
    max_tokens=512,
)
print(resp.choices[0].message.content)

The API is stateless — you must resend the conversation history with each request. The web chat and mobile app keep session history for you; the API does not. Both V4 tiers support a 1,000,000-token context window with output up to 384,000 tokens, and thinking mode is a request parameter on either model (reasoning_effort="high" with extra_body={"thinking": {"type": "enabled"}}), not a separate model ID. When thinking is enabled, the response returns reasoning_content alongside the final content.

Compliance posture: who should avoid the hosted API

What this means for cost and planning

Teams that can use the hosted API still get the price advantage that has driven DeepSeek’s US adoption. V4-Flash lists $0.028 cache-hit / $0.14 cache-miss / $0.28 output per million tokens; V4-Pro lists $0.145 / $1.74 / $3.48. Our DeepSeek API pricing breakdown runs worked examples for both tiers. Teams that need to self-host for compliance reasons should model GPU costs with the DeepSeek hardware calculator before committing.

How US restrictions compare internationally

The US pattern — device-level bans, no consumer ban — mirrors what most allies have done. The Australian government blocked access to DeepSeek on all government devices on national security grounds, Taiwan took similar action, and the Italian data protection authority announced limitations on the processing of Italian users’ data by DeepSeek. South Korea’s industry ministry also temporarily blocked employee access to the app. Italy has been the outlier, pursuing a data-protection enforcement action rather than a device rule. For the full country-level picture, see DeepSeek availability by country.

What to watch next

1. Federal bill progress. HR 1121 and S 765 remain in committee. If either clears and is signed, OMB would have 60 days to issue removal guidance across executive-branch IT.
2. FAR Council rule-making. A formal Federal Acquisition Regulation amendment would extend the reach from federal employees to every federal contractor.
3. State attorneys general. The 21-AG letter to Congress suggests more state actions are likely, particularly in states that followed the TikTok precedent closely.
4. Hawley-style export/import controls. A broad US–China AI decoupling bill, if advanced, would affect commercial users too, not just government ones.
5. Court challenges. No significant legal challenge to a DeepSeek state ban has reached public dockets as of April 2026; that may change as contractor-facing rules appear.

For ongoing coverage, our DeepSeek news hub tracks policy changes as they happen, and the DeepSeek latest updates feed carries shorter interim notes.

Verdict

Calling this situation a “ban” overstates it. The DeepSeek US restrictions in effect as of April 24, 2026 are targeted at a defined surface — government devices, some federal agency networks, and (if the Cassidy–Rosen bill passes) federal contractors at work. Private developers, researchers, and companies outside those categories remain free to use DeepSeek. The prudent move for sensitive workloads is to lean on the open-weight path — self-hosting V4-Flash or V4-Pro on US infrastructure sidesteps the cross-border data issue that motivated every one of the restrictions catalogued above.

Last verified: 2026-04-24. DeepSeek AI Guide is an independent resource and is not affiliated with DeepSeek or its parent company. Model IDs, pricing and API behaviour change; check the official DeepSeek documentation and pricing page before committing to a production decision.