Dynamo AI announced the launch of Foundation Guardrails designed to help organizations comply with Article 5 of the European Union's Artificial Intelligence Act (AI Act), which addresses Prohibited AI Practices. Dynamo AI's new guardrails deliver controls for text-based AI systems to prohibit practices the law has deemed to have an "unacceptable risk" and that go into effect on February 2, 2025. As Foundation Guardrails, these guardrail models can be used out-of-the-box or customized to meet enterprise requirements.

The AI Act is a landmark piece of legislation that will regulate the use of AI in the European Union. It introduces a risk-based approach to AI regulation, with different requirements for different types of AI systems. While the new DynamoGuard Foundation Guardrails are designed to mitigate risk and facilitate compliance with the AI Act, they are also suited for general use to protect against harmful AI outcomes and promote responsible AI.

"The EU AI Act is extraterritorial so it does not just impact European-based organizations . It's important for enterprises conducting business in the EU to have a host of effective and demonstrable controls to support compliance with the law, which is where our new Foundation Guardrails become critical," said Daniel Ross, Head of AI Compliance Strategy at Dynamo AI. "A range of global enterprises are  subject to compliance with the EU AI Act, and we are here to support them as they continue to use responsible AI across borders."

DynamoGuard's EU AI Act guardrails prohibit:

• Manipulation or deceptive language (output guardrail)
• Social scoring (input and output guardrails)
• Criminal offense risk assessment (input and output guardrails)
• Biometric inference or categorization (input and output guardrails)
• Emotion recognition (output guardrail)

Dynamo AI's new Foundation Guardrails help enterprises not only put responsible AI guardrails in place, but also help them avoid the significant financial penalties that come with non-compliance. According to the AI Act, non-compliance with the prohibition of the AI practices referred to in Article 5 shall be subject to administrative fines of up to 35 000 000 EUR (more than $36,000,000) or, if the offender is an undertaking, up to seven percent of its total worldwide annual turnover for the preceding financial year, whichever is higher.

Dynamo AI's new Foundation Guardrails for the AI Act are available now in the Dynamo Platform.