How do I get access to Gemini 4 Argon?

As of October 1, 2026 there is one route: apply to Google DeepMind's Fairwind Program at deepmind.google/fairwind-program. It is for high-priority defenders — governments, healthcare providers, telecoms, security vendors — and only their internal cybersecurity, incident-response or penetration-testing teams may be granted Argon access. Google runs background checks on applicants' security history and ethical record. Everyone else waits for the next phase: paid Gemini API customers and Google AI Ultra subscribers, which Google says will come 'as soon as possible' with no date.

Who qualifies for the Fairwind Program?

Organisations Google classes as high-priority defenders: governments, healthcare providers, telecommunications services, cloud customers and security partners such as Wiz. Applicants must pass Google's due-diligence check of their security history and ethical operations, agree to user-level authentication, phishing-resistant MFA and access controls, restrict Argon to internal security teams, track every employee's access and use, and never share, redistribute or sell access. Individuals and general developers do not qualify.

What can Fairwind partners do with Gemini 4 Argon?

Defensive and research work only. Partners receive Argon without its cyber guardrails for vulnerability discovery, validation and patching, plus the CodeMender code-security agent. Permitted dual-use tasks are authorised threat simulation, reverse engineering and malware analysis for defensive or academic purposes. Malware creation and offensive use are prohibited, and misuse of the privileged access is grounds for removal.

When will Gemini 4 Argon be publicly available?

Google has not given a date. Its September 30, 2026 announcement says Argon is in the U.S. government's voluntary pre-release model access process and that Google will 'gradually expand access' while iterating on guardrails, with developers, enterprises and consumers reached 'as soon as possible, starting with paid API customers and Google AI Ultra subscribers.' DeepMind's Koray Kavukcuoglu said in late September that Google wanted Gemini 4 out well before year-end, so a Q4 2026 API rollout is the reasonable expectation, not a promise.

Quick Answer

How to Get Gemini 4 Argon Access: Fairwind Program Guide

Published: October 1, 2026

The short answer

As of October 1, 2026, the only way to use Gemini 4 Argon is through Google DeepMind’s Fairwind Program, which admits vetted cyber-defence organisations after a background check and restricts the model to their internal security teams. Paid Gemini API customers and Google AI Ultra subscribers are explicitly next in line; Google has given no date. Below is the eligibility checklist, what the application commits you to, and what to do if you do not qualify. Facts verified October 1, 2026.

Step 1: Check whether you are a “high-priority defender”

Fairwind’s stated audience is organisations whose compromise would hurt the public: governments, healthcare providers, telecommunications services, critical-infrastructure operators, cloud customers and security vendors. Wiz is the named launch partner, using Argon in its Scan for Good programme to find exposures in public infrastructure for free.

If you are a solo developer, a startup without a security function, or a company that wants Argon for coding or research, you do not qualify — Fairwind only lets partners grant Argon access to “internal cybersecurity, incident response, or penetration testing teams.”

Step 2: Prepare for due diligence

Google says it “conducts background checks on organisations that apply, to verify security history and analyze their record of ethical operations.” Have ready:

A description of your defensive mandate and the systems you protect
Evidence of security posture: user-level authentication, phishing-resistant MFA, access controls
The named team(s) that would receive access and how you will track each employee’s use
Confirmation you will not share, redistribute or sell access

Step 3: Apply

Fill in the form at deepmind.google/fairwind-program (“Apply for access”). Google says it will “review and respond to eligible partners who meet our criteria as soon as we can.” No SLA is published. The programme launched September 2, 2026 with Gemini 3.8 Flash Cyber; Argon was added September 30, so expect a queue.

Step 4: Understand what you get — and what is off-limits

Accepted partners receive:

Gemini 4 Argon without cyber guardrails — the full vulnerability-detection and automated-patching capability. Google says this is the same configuration its own internal security teams use.
CodeMender, DeepMind’s code-security agent that finds, reasons about, validates and emits deployment-ready fixes without you building your own harness.
Likely Gemini 3.8 Flash Cyber as well, the earlier Fairwind model.

Permitted: authorised threat simulation, reverse engineering, malware analysis for defensive and academic research. Prohibited: malware creation, offensive operations, any use outside the named security teams. Argon’s CBRN refusals remain in place; only the cyber guardrails are lifted.

If you do not qualify: what to do now

Build on Gemini 3.8 Flash today ($0.75/$3.75 through December 31, 2026) and design a model-tier switch so Argon can slot in for the hardest tasks. See Argon vs 3.8 Flash.
Get on a paid Gemini API plan or Google AI Ultra. Google names both as the first public cohort. There is no waitlist to join; being a paying customer is the waitlist.
Budget at $4/$20, not $2/$10. The intro price has no end date; Vals already lists Argon at $4/$20.
Benchmark the alternatives. Claude Opus 5.5 ($4/$20) beats Argon on Terminal-Bench 4.0; GPT-6 Astra ($10/$50) beats it on FrontierSWE v2. See Argon vs Opus 5.5 vs Astra.
Need gated cyber capability now? OpenAI’s Daybreak programme (GPT-5.6 Cyber) and Anthropic’s Cyber Verification Program (Mythos 5.1) run parallel vetting tracks — compared here.

Why Google is gating it

Google’s post says releasing “frontier capabilities at this level requires a phased approach”: it is in the U.S. government’s voluntary pre-release access process, strengthening misuse refusals, prompt-injection defences (0.7% Gray Swan IPI success rate), chain-of-thought misalignment monitoring and sealed sandboxes before broad release. The timing is not accidental: OpenAI cancelled GPT-6.1 Astra on September 28 over safety findings, the White House AI accord was signed September 29, and the FTC confirmed a probe of OpenAI and Anthropic on September 30. A defenders-first rollout is Google’s answer to all three.

Last verified: October 1, 2026.