IBM today announced the upcoming launch of watsonx.governance, a new offering designed to help organizations responsibly govern and scale generative AI systems like large language models (LLMs). With capabilities to manage risk, ensure transparency, and anticipate compliance, watsonx.governance aims to address growing concerns around AI fairness, bias, and explainability.
The rise of powerful generative AI has unlocked new possibilities for businesses to enhance services and workflows. However, these data-hungry models also introduce potential risks, as training data may contain biases and lack clear provenance. Furthermore, LLMs can produce outputs difficult for humans to interpret. This opaqueness makes it challenging for organizations to fully trust AI decision-making.
As Katrina Troughton, IBM Fellow and CTO AI Applications and Outcomes, explained, "Businesses want to tap into the possibilities of generative AI, but struggle with the lack of transparency and inability to properly govern these models."
watsonx.governance seeks to provide that missing transparency with capabilities to document datasets, models, and AI pipelines. This traceability helps organizations better explain AI reasoning. The offering also monitors models for fairness, bias, and drift issues, enabling preventative and corrective actions.
With impending regulations on AI worldwide, watsonx.governance additionally helps enterprises translate policies into enforceable governance. Customizable dashboards and reporting provide visibility into AI risks and compliance across the business.
IBM Consulting has expanded its services to assist clients in scaling responsible AI governance across technology, processes and people. This human-centric approach complements watsonx's focus on explainable, ethical AI.
As Kareem Yusuf, SVP of Product Management and Growth for IBM, stated, "Its ability to translate regulations into enforceable policies will only become more essential for enterprises as new AI regulation takes hold worldwide."
watsonx.governance integrates with models from all sources, including LLMs from IBM, partners, and open source. Along with the rest of the new watsonx platform, it aims to make enterprise AI adoption more transparent, responsible, and ultimately, trustworthy.