OctoAI Safety Center

Safety Principles

At OctoAI, safety is a top priority. We work hard to prevent our products and services from being used in ways that harm individuals or society. To work toward this goal, we developed a suite of policies and tools to enhance the safety of the OctoAI platform, and allow customers to leverage these for their own applications. We partner with our customers and industry leaders to assure that we have a layered approach to protection.

This work is never done. We expect our Trust & Safety practices to evolve as OctoAI grows and develops. We continually seek to learn and adapt our program to keep up with the rapid pace of AI innovation. As part of our program, we want our customers to understand how to safely use OctoAI and how OctoAI can help secure their own products. This center provides more information about our policies and practices.

What is not allowed

We want OctoAI’s platform and associated demos and services to be used safely and responsibly. We have created usage policies that apply to all OctoAI users. To that end, you are responsible for complying with our Acceptable Use Policy and our Terms of Use.

We do not allow any of our products or services to be used to enable, conduct, or incite any illegal or harmful activities, including:

• Child sexual exploitation or abuse

• Sexual exploitation or human trafficking

• Scams, malware or plagiarism

• Harassment or bullying

• Violence or terrorism

• Hate or discrimination

• Disinformation or spam

• Non-consensual intimate imagery or sextortion

• Dangerous and regulated goods

• Unauthorized copyright access

OctoAI is a platform that developers can use to build their own end user applications, and developers who do so are responsible for ensuring their users who may interact indirectly with our service are also abiding by these usage policies. For a full review of what is not allowed, please review our Acceptable Use Policy.

Safety features

Our team is committed to developing and leveraging safety features to prevent, monitor and report unsafe use of our products, code samples, and demo apps. We actively collaborate with customers, external partners and industry leaders to improve our understanding of the issues and enhance our safety features.

Prevention

OctoAI has developed several product safety features to prevent and block harmful content. These features include prompt filters to block any unsafe prompts from generating content and an image output classifier to mitigate the generation of unsafe content. Llama Guard will soon be available for customers to incorporate into their own text-gen applications and workflows.

In building our safety features, we follow our Acceptable Use Policy and Terms of Use to determine what content is unsafe.

Monitoring

We proactively monitor and test our safety filters to understand how individuals may try to bypass our protections and to improve the efficacy of these filters. We also work directly with our customers to help them identify gaps in their own protections and to ensure they are meeting our safety requirements.

Reporting

OctoAI uses various methods to determine whether users are following these policies. We also encourage and appreciate user reports of violations, suspected violations, or unexpected model behavior. Reports may be sent to safety@octo.ai.

When appropriate, OctoAI will report illegal or harmful behavior to the relevant government authorities.

FAQ

How do I find out if a certain behavior is allowed at OctoAI?

We try to be as explicit as we can in the behaviors that are not allowed at OctoAI. If a behavior is illegal, it is never allowed. If you are unsure, please start by reviewing our Acceptable Use Policy. After you have reviewed our policy, you can reach out if you still have questions to safety@octo.ai.

What happens if we violate the Acceptable Use Policy?

If we discover that your product or usage is violating our usage policies, we may take any of the following actions:

• Ask you to make changes to your product or usage while continuing to use OctoAI

• Suspend your account pending changes to your product or usage

• Terminate your account

• Issue a report to the relevant government authorities

How do I report policy violations or safety concerns?

Users are encouraged to report unsafe behavior, suspected violations, or unexpected model behavior. We accept reports at safety@octo.ai.

When you submit a report, it gets sent immediately to our Trust & Safety team for review. If the content is deemed to violate one of our policies, we will take further actions.

I was removed from OctoAI's platform. How do I appeal this decision?

We welcome all users to appeal actions taken against their account if they believe there was a mistake. Submit the details surrounding your removal to safety@octo.ai. We are happy to review your situation. However, we take violations seriously, and we will not allow content or behavior that violates our policies on our platform.

How are you protecting my data?

Keeping user data private and secure is our priority. We continually invest in security capabilities and practices in our platform and processes. OctoAI uses your Customer Data solely to deliver our service to you. This may include periodic, audited access for debugging or safety monitoring. However, we do not use your data to train models–unless you direct it as part of our fine-tuning features–and we never sell it to third parties.

OctoAI is also SOC II Type 2 compliant. For more information on our security and privacy procedures, please review our relevant policies:

• Privacy Policy
• Data Privacy Framework Certification
• Security
• Responsible Disclosure