Gå til hovedindhold
OpenAI

Anmod om OpenAI Pilot: Trusted Access For Cyber

Trusted Access for Cyber gør det muligt for godkendte virksomhedskunder og cybersikkerhedsspecialister at bruge vores mest kapable modeller til cybersikkerhedsarbejde med dobbelt anvendelse. Vi betragter disse kapaciteter som stærke kraftmultiplikatorer for netværksforsvarere, samtidig med at vi erkender, at ondsindede aktører kan forsøge at udnytte de samme værktøjer til at øge omfanget og sofistikeringen af deres operationer.

Trusted Access for Cyber reducerer denne risiko ved at kræve, at medlemmerne giver yderligere identifikation og oplysninger om professionel brug. Selvom disse foranstaltninger ikke forventes at forhindre alt potentielt misbrug, virker de i kombination med vores eksisterende cybersikkerhedsforanstaltninger for at reducere risikoen for skade betydeligt, hvilket gør det muligt for os at tilbyde kapaciteter med højere risiko og større effekt for et bredere fællesskab af forsvarere.

Krav for at blive medlem:

  • Nøjagtig udfyldelse af denne ansøgning i sin helhed. Unøjagtige eller ufuldstændige oplysninger kan medføre forsinkelser eller manglende accept
  • Villighed til at give yderligere oplysninger eller afklaringer, som OpenAI anmoder om, før og efter adgang

Oplysninger om kunde/enhed

A. Identifikation af organisation

B. Primær kontaktperson (den person, der indsender)

C. Forhold til OpenAI

Professionelt brugsscenarie

D. Hvordan har du til hensigt at benytte Trusted Access til Cyber?

Juridiske vilkår og attestationer

Trusted Access for Cyber Participant Addendum

These Trusted Access for Cyber Terms (these “Terms”), together with the intake responses and attestations submitted through the intake form at https://openai.com/form/enterprise-trusted-access-for-cyber/⁠ or attached hereto (the “Intake Form” and collectively with the Terms, the “TAC Addendum”), form part of the Services Agreement between OpenAI and the entity named in the Intake Form (“Customer”), and govern Customer’s access to models made available to Customer through the Trusted Access for Cyber (“TAC”) program. "Services Agreement" means the OpenAI Services Agreement available at https://openai.com/policies/services-agreement/⁠⁠ unless and to the extent that (a) Customer and OpenAI have signed an Enterprise Agreement for OpenAI Services, in which case such agreement will govern, or (b) Customer  will access TAC on Amazon Bedrock, in which case the “OpenAI Services Agreement - Amazon Bedrock” available at https://aws.amazon.com/legal/bedrock/third-party-models/⁠(åbner i et nyt vindue) will govern. To participate in TAC, Customer must submit the information in the Intake Form and receive approval from OpenAI. 

  1. Representations and Warranties. Customer represents and warrants that the information in the Intake Form is true and correct in all respects, and Customer will notify OpenAI if there are any material changes to the information submitted in the Intake Form during the term in which TAC is used. Customer acknowledges and agrees that provision of false or misleading information in the Intake Form, or any failure to notify OpenAI in writing if any information provided in the Intake Form has changed or is no longer complete and accurate, will constitute a material breach of the TAC Addendum and the Agreement.

  2. Approved Use Cases. Participation in TAC comes with heightened responsibility. Access to advanced cyber capabilities is granted only for legitimate, defensive, and authorized security purposes (“Approved Use Cases”). Approved Use Cases are intended to enable real-world security testing, vulnerability research, and defensive readiness and do not include uses that may cause harm, disruption, or unauthorized access. Customer is only authorized to use TAC to enable its own personnel to engage in Approved Use Cases and only in compliance with OpenAI’s usage policies at: https://openai.com/policies/usage-policies/⁠, as well as OpenAI’s Cyber Abuse Policy.

  3. Cyber Abuse Policy. We disallow use of our Services to facilitate Cyber Abuse. “Cyber Abuse” means unauthorized access, exploitation, credential theft, data exfiltration, malware or destructive capabilities, social engineering, evasion, lateral movement, denial-of-service activity, or assistance to any sanctioned entities or identified malicious cyber actors.  This does not prohibit benign defensive, educational, research, privacy-protective, incident response, or authorized security testing uses, including malware or vulnerability analysis, responsible disclosure, and red-team planning, so long as the activity does not enable real-world harm, target live systems without authorization or provide actionable assistance for abuse.

  4. Approved Access Credentials. “Approved Access Credentials” are, collectively, the Org ID, API keys, and specific end user IDs approved by OpenAI for any Approved Use Case. Customer may only make TAC access available to its personnel with a valid need to access TAC, and Customer is responsible for all activities that occur using the Approved Access Credentials. Customer may not (a) share or otherwise make the Approved Access Credentials available to third parties, or (b) share or sell access to the capabilities provided under the Approved Access Credentials to third parties without OpenAI’s express prior written consent. Customer must immediately notify OpenAI if it becomes aware of or reasonably suspects unauthorized access or use of Approved Access Credentials, including any use of TAC for a purpose that is not an Approved Use Case.

  5. Additional Actions. OpenAI will have the right to take any action it deems necessary, in its sole discretion, to preserve the safety, security and integrity of the Services or its business, including, without limitation, suspending or terminating Customer’s access to TAC without refund, requiring additional documentation or assurances to permit continued TAC, conditioning TAC use on Customer’s acceptance of additional requirements or limitations, and pursuing all legal and equitable remedies permitted by applicable laws.

  6. Indemnification. Notwithstanding anything to the contrary in the Services Agreement, unless Customer is a government entity prohibited by law from agreeing to this Section, Customer will indemnify, defend, and hold OpenAI and its affiliates harmless against any liabilities, damages, and costs (including reasonable attorneys’ fees) payable to a third party arising out of any use of the Approved Access Credentials in violation of this TAC Addendum. Limitations on Customer’s liability in the Services Agreement do not apply to this TAC Addendum.

If OpenAI approves you for Government Trusted Access for Cyber ("GTAC"), the following additional terms apply.

  1. GTAC access is intended only for approved users performing lawful, authorized defensive cybersecurity work supporting an approved government mission, government environment, or government-validated protected infrastructure. 

  2. Consistent with the OpenAI Usage Policies, GTAC may not be used for offensive operations, gaining unauthorized access, malware deployment or improvement, credential theft, phishing, data exfiltration, sabotage, or denial-of-service activity.

  3. GTAC is not available for general corporate security, commercial product development, resale, proxying, embedding, customer-facing product traffic, downstream third-party use, or other work outside the approved GTAC scope. Contractors, defense industrial base companies, integrators, cybersecurity vendors, and other non-government entities may use GTAC only to support the approved GTAC scope.


  1. Jeg har gennemgået Ansøgningsformularen og erklærer og bekræfter, at de angivne oplysninger er korrekte og fyldestgørende efter min bedste viden og overbevisning efter behørig undersøgelse og rimelig omhu.

  2. Jeg har den nødvendige bemyndigelse til at underskrive denne formular på vegne af kunden og har indhentet alle interne godkendelser, der kræves for at overholde TAC.

  3. Jeg vil straks underrette OpenAI skriftligt, hvis jeg opdager, at nogen oplysninger i denne Ansøgningsformular ikke er, eller ikke længere er, korreke og fyldestgørende.

  4. Kunden accepterer de juridiske vilkår i denne Ansøgningsformular, som hermed indarbejdes i Tjenesteaftalen.


Bemærk: Efter indsendelse skal du gennemføre et kort bekræftelsestrin for at bekræfte din identitet, herunder en kontrol af et offentligt udstedt ID og angivelse af grundlæggende virksomhedsoplysninger.