Safety

AI needs to be useful and accessible. That's why it needs to be safe to collaborate with agents and connect them to real world systems. We adopted a multi-facted approach to safety and reliability:

• Monitoring
• Universal Accessibility
• Global Threat Detection
• Automatic Testing
• Stability
• Privacy
• Sharing
• Restricted and controlled regions
• Law enforcement

Do you have ideas about this? Chat with us over at [email protected]

Monitoring

With monitoring of agent reasoning and planned actions, we can prevent potential safety risks and ensure agents operate as intended. This can be used to block agent actions before they happen, and adjust behaviour through training.

Universal Accessibility

AI is for everyone, therefore it should be possible to train personalised agents that fit with unique needs, cultural perspectives, languages and customs while also meeting global performance standards. It's our mission to make this possible through AI systems on the Niyeo platform.

Global Threat Detection

As agents become more capable and valuable, so can they attract more possible threats. On Niyeo, such threats are detected and blocked on a global platform, sometimes before they happen. Everyone who works with agents or trains agents on Niyeo benefits from this shared protection.

Automatic Testing

Automated testing saves time, while ensuring agents work as intended. Not having to test agents manually for every aspect, means agents can be tested regularly and in controlled environments. This allows agents to be compared against benchmarks, and check if their performance improves over time.

Stability

Niyeo exclusively uses custom trained, fine tuned or open source models hosted in multiple datacenters. By reducing reliance on a handful of closed AI providers, we can offer high stability and reliability.

Privacy

By not using any third party AI providers, we can ensure that user data is not used for training or other purposes. And reduce exposure of private data to third party AI providers.

Research Community

A community of researchers around the world co-create and openly share ideas, open-source tools and data to improve AI. Niyeo also aims to contribute to this community to foster collaborative relationships. That benefits people who train agents on Niyeo, through implemnentation of the latest methods in AI safety. It creates the opportunity for Niyeo to work with the top talent in the field of AI.

Restricted and controlled regions

Niyeo has limited service availability in areas with active conflicts or sanctions.

Information for Law Enforcement Authorities

People and AI worlwide use Niyeo to communicate and collaborate. We cooperate with law enforcement agencies to maintain user safety, reviewing and responding to requests as per applicable laws and policies. Law enforcement officials seeking records from Niyeo should follow the guidelines below. For additional questions, please contact us directly. Do not send law enforcement inquiries through Niyeo Support or other unintended channels.