OpenAI Updates Usage Policies: Key Changes for June 2026
OpenAI Updates Usage Policies: Key Changes for June 2026
OpenAI has updated its Usage Policies, Service Terms, and Services Agreement, effective June 2, 2026. These revisions clarify user responsibilities, reinforce that AI output is not a substitute for professional advice, and detail guidelines for new features like GPTs, Visual Capabilities, and ChatGPT Sites.
What are the main changes in OpenAI’s latest policies?
OpenAI’s latest policy updates, effective June 2, 2026, encompass revisions to their Service Terms, Terms of Use, and Services Agreement. A primary focus of these changes is a clear reiteration that AI-generated content from their tools, including ChatGPT and API services, should not be considered a replacement for professional guidance in fields such as medicine, law, or finance. This move aims to mitigate liability and emphasize the need for human oversight and judgment when utilizing AI in critical decision-making processes. The policies also introduce detailed frameworks for managing new features like custom GPTs, advanced Visual Capabilities, and the recently launched ChatGPT Sites.
Specifically, the updated Service Terms clarify OpenAI’s indemnification obligations for API customers concerning intellectual property infringement claims related to AI output. It outlines exceptions where this indemnity does not apply, particularly when users knowingly generate infringing content, disable safety features, or use output with unapproved modifications. For Enterprise customers, similar indemnification details are provided, highlighting the shared responsibility in ensuring ethical and legal use of AI.
The indemnity carve outs are particularly noteworthy for organizations deploying AI at scale. The six enumerated exceptions cover scenarios ranging from known infringement risks to trademark violations in commercial contexts. This effectively shifts significant liability onto API users who push the boundaries of acceptable use. For example, any modification or transformation of AI output combined with other products creates a scenario where OpenAI’s IP protection falls away entirely, placing the burden squarely on the developer to ensure their integrations do not violate third-party rights.
The revised Usage Policies reinforce OpenAI’s commitment to safety and responsibility. They consolidate rules across all OpenAI products and services, emphasizing universal prohibitions against harmful uses such as threats, harassment, illegal activities, and the unauthorized use of private information. The policies also address the sensitive area of “Medical Use,” stating unequivocally that services are not intended for diagnosis or treatment and users bear responsibility for compliance with medical regulations. Additionally, new sections detail guidelines for Image and Video Capabilities, particularly regarding facial recognition and the reproduction of likenesses, and the use of Apps and Actions within ChatGPT for interacting with third-party applications.
How do the new policies affect API users?
For developers and businesses leveraging OpenAI’s API, the updated Service Terms (June 2, 2026) bring important clarifications regarding intellectual property (IP) indemnity. OpenAI now specifies that its indemnification covers third-party claims of IP infringement stemming from the use or distribution of AI-generated output, provided certain conditions are met. However, this indemnity has crucial limitations. It does not apply if API users or their End Users knew or should have known the output was infringing, disabled or ignored safety features, or if the output was modified or used in combination with non-OpenAI products. This underscores the need for API users to integrate OpenAI’s tools responsibly and diligently, ensuring adherence to documentation and internal checks for potentially infringing content.
Furthermore, the policies make it clear that developers must adhere to the applicable documentation at OpenAI’s platform docs. This is a standard but important reminder for continuous compliance. The explicit mention of code generation output potentially being subject to third-party licenses, including open-source licenses, specifically impacts those using features like OpenAI Codex. Developers are advised to be aware of and comply with any such licensing terms associated with the generated code.
The updated App Developer Terms are also relevant for those creating custom Apps and Actions that interact with ChatGPT. These terms govern the responsibilities for ensuring compliance, particularly when an App sends user data to an external Application. OpenAI maintains the right to review, refuse, or remove any GPT or App for various reasons, including legal or security concerns, or policy violations. This means developers must build with robust ethical considerations and constant policy adherence in mind to avoid disruption.
What are the implications for professional advice?
A significant highlight of the updated policies is OpenAI’s explicit stance that its AI tools are not substitutes for professional medical, legal, or other regulated advice. While users can still query generative AI models for information on these topics, the output should not be relied upon as definitive professional guidance. This revision aims to limit OpenAI’s liability and reinforce the necessity of human expertise and discretion, especially in professions where accuracy and accountability are paramount. Law firm Baker Donelson highlighted that these changes serve as a critical reminder for organizations deploying AI to maintain subject matter expertise and human oversight.
The Usage Policies (effective October 29, 2025) strictly prohibit the provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional. This has profound implications for businesses and individuals who might have previously used AI to generate such advice. It necessitates a re-evaluation of workflows and increased scrutiny over AI-generated content in sensitive areas. Companies that use OpenAI’s models in healthcare, finance, or legal tech applications must ensure their processes include licensed professionals validating and contextualizing AI outputs to meet legal and ethical standards.
This clarification also extends to the “Medical Use” section of the Service Terms, stating that OpenAI services are not intended for diagnosis or treatment. This directly impacts any application attempting to use AI for medical imaging analysis, signal acquisition systems, or in vitro diagnostic devices. Users are reminded of their responsibility to comply with all applicable laws and regulations when using AI in a medical or healthcare context, further reinforcing the boundary between AI assistance and professional liability.
The healthcare implications extend beyond individual practitioners to entire institutional workflows. Hospitals, telemedicine platforms, and health-tech startups that have integrated ChatGPT or API services into patient interaction pipelines must now review those implementations against the updated policies. The explicit carve out for signal acquisition systems and ECG analysis, for instance, suggests OpenAI is drawing a firm line around diagnostic applications that were already under scrutiny from regulators. Organizations in this space will need to document their human oversight processes more rigorously and may need to seek alternative AI tools specifically designed and certified for medical use.
How do the policies address new features like GPTs and Visual Capabilities?
OpenAI’s updated Service Terms and Usage Policies provide specific guidelines for newer features, reflecting their evolving ecosystem. For custom GPTs, distinct policies are outlined for both “Builders of GPTs” and “Users of GPTs.” Builders are held solely responsible for their GPT Content (name, instructions, description), Actions, and configurations. They must ensure their GPTs comply with both the main Agreement and the Usage Policies. Critically, by sharing a GPT, builders grant OpenAI a broad license to use, distribute, and promote their GPT, including its content. This indicates that while GPTs offer customization, they operate within a framework where OpenAI retains significant control over distribution and usage.
For users of GPTs, the policies clarify that most GPTs are created by other users and may rely on third-party applications not controlled by OpenAI. This means users must exercise caution and “only use GPTs that you know and trust.” The platform also provides abuse reporting features for GPTs that violate policies. Similarly, for Apps and Actions connecting ChatGPT to third-party applications, users are solely responsible for actions taken and must trust the underlying application and its privacy policies before enabling.
Regarding Visual Capabilities, which allow models to accept images and videos as input (e.g., for Sora or for multimodal agents), the policies introduce strict prohibitions. These include not using Visual Capabilities to identify individuals, solicit private information, or reproduce someone’s likeness without explicit consent. When sharing images or videos publicly on the Services, users grant OpenAI rights to reproduce, distribute, and modify the content for operating and promoting the Services, with limited rights for other users to remix the content solely within the platform. These rules reflect a proactive effort to address privacy and consent concerns associated with advanced AI capabilities.
The licensing terms for user-shared content on services like Sora deserve particular attention. By uploading a cameo or sharing an image, users grant OpenAI a worldwide right to reproduce, distribute, modify, and display that content for promoting the Services. This is standard for most user-generated content platforms, but the specific inclusion of cameo images used in AI-generated videos means users should be deliberate about what they share publicly. The limited remix rights granted to other users apply strictly within the platform, preventing external misuse while enabling creative collaboration in the AI-generated content ecosystem.
What is the effective date of OpenAI’s updated policies?
The OpenAI Service Terms were updated on June 2, 2026, and went into effect immediately. The overarching Terms of Use were published on January 1, 2026, and also became effective on that date. The Usage Policies received a significant update on October 29, 2025, shifting toward consolidated “Universal Policies” across all OpenAI products. According to LLM Stats, this June cycle marks the latest in a series of quarterly policy refreshes that have steadily tightened usage terms as AI adoption has expanded across regulated industries.
Key Takeaways:
OpenAI’s updated policies (effective June 2, 2026) differentiate AI outputs from professional advice, emphasizing human oversight.
API users face clarified IP indemnification terms with specific exceptions related to responsible usage and safety features.
New guidelines for GPT builders and users stress accountability for content, actions, and the responsible selection of trusted GPTs.
Visual Capabilities are subject to strict rules on privacy, consent for likeness reproduction, and data use to prevent misuse.
The policies emphasize a universal, consolidated approach to responsible AI use across all OpenAI products, with strong prohibitions against harmful activities.
Frequently Asked Questions (FAQs)
Q1: Are OpenAI’s AI models now liable for incorrect professional advice they provide?
A1: No. OpenAI’s updated policies explicitly state that its AI tools, including ChatGPT and API services, are not substitutes for professional medical, legal, or other regulated advice. While users can still query the models for such information, they are ultimately responsible for seeking professional counsel and exercising human judgment, as the AI output is not intended to be a definitive source of regulated advice.
Q2: What happens if I disable OpenAI’s safety features?
A2: Disabling, ignoring, or not using any relevant citation, filtering, or safety features provided by OpenAI can negate OpenAI’s indemnification obligations regarding third-party intellectual property claims. This means you could be held solely responsible for any resulting legal issues from the AI’s output. It is crucial to use all provided safety and citation tools.
Q3: Can I use OpenAI’s Visual Capabilities for facial recognition?
A3: No. OpenAI’s policies strictly prohibit the use of Visual Capabilities to assist in identifying a person or to solicit or infer private or sensitive information about an individual. Reproducing someone’s likeness without explicit consent and all necessary rights is also prohibited, highlighting strong privacy and ethical guidelines for image and video processing.
Q4: Do the new policies apply to custom GPTs I’ve built?
A4: Yes. Both “Builders of GPTs” and “Users of GPTs” are subject to specific guidelines within the updated Service Terms and Usage Policies. Builders are responsible for their GPT content and ensuring compliance, while users are advised to only engage with GPTs they know and trust, as many are created by third parties and not controlled by OpenAI.
Q5: Where can I find more analysis about these policy changes?
A5: For detailed legal analysis, the Baker Donelson article provides excellent coverage of the implications for organizations deploying AI. Additional context on evolving AI model usage terms can be found on LLM Stats’ AI Trends page and amstlegal’s breakdown of ChatGPT terms for businesses.