IBM unveils watsonx.governance to reassure businesses they can trust generative AI

IBM has announced a tool to help businesses build trust in generative AI. Watsonx.governance comes online in December, with the promise that it will improve transparency and “explainability”.

The product is already in Tech Preview, and interested parties can book a live demo via the website. IBM promises that the tool will give “company boards and CEOs” the reassurance that may be holding them back from introducing generative AI.

“Watsonx.governance is a one-stop-shop for businesses that are struggling to deploy and manage both LLM and ML models, giving businesses the tools they need to automate AI governance processes, monitor their models, and take corrective action, all with increased visibility,” said Kareem Yusuf, Senior Vice President, Product Management and Growth, IBM Software in a statement.

“Its ability to translate regulations into enforceable policies will only become more essential for enterprises as new AI regulation takes hold worldwide.”  

What will watsonx.governance do?

IBM announced four key elements to watsonx.governance:

• The ability to monitor new large language model (LLM) metrics
• Increased transparency and “explainability”
• Validation tools for LLMs
• Better monitoring of a model’s health details such as data size, latency, and throughput to identify bottlenecks, compute intensive workloads, and drift of text models

The fundamental idea here is to avoid the black box nature of current generative AIs. You put in a prompt and get an answer, whether that’s words, images, code or anything else. But how was that data trained? Is the reasoning transparent? What happens if it’s simply wrong?

Watsonx.governance is meant to give businesses (and governments) a “toolbox” that allows them to put protections in place and have trust in their generative AI system.

What do CEOs and CTOs think about watsonx.governance?

Given that IBM is targeting CEOs and board members with watsonx.governance, we asked a selection for their views.

“Many business leaders are increasingly concerned about the transparency and explainability of AI systems,” said Julien Salinas, CEO at NLP Cloud, an AI platform. “As decision-making processes become more reliant on AI, understanding how models work and ensuring they are free from bias is critical. A governance tool like Watsonx.governance would likely be seen as a step in the right direction.”

However, Julien has reservations. “I am dubious about the concrete applications of such a platform though. LLM ‘explainability’ has been a hot topic for a couple of years now, and for the moment nobody has really solved the problem. So I definitely do not consider such a measure as a ‘silver bullet’, but it is a step in the right direction.”

Vaclav Vincalek, the virtual CTO and founder of 555vCTO, echoes these thoughts. “CEOs and CFOs of any organisation are responsible for managing risk. Having a tool which can help them create a governance framework around a technology that the rest of their organisation wants to use is a must.”

He adds that as the technology matures, so will the regulations. “Eventually it will be incorporated in compliance frameworks such as SOC2 and it will become a part of a company’s ‘business as usual’ mantra.”

In conclusion: “The use of tools like watsonx.governance will force everyone involved to think a little bit harder about the data and what it’s going to be used for, as well as the ominous ‘what could go wrong’ scenario.”

Will watsonx.governance allay company’s fears?

Jignesh Patel is a computer science professor at Carnegie Mellon University and co-founder of DataChat, a no-code platform that lets you “chat with your data”. Although he’s fully behind the need for governance, he believes watsonx.governance is a first step rather than a complete solution.

“It will help, but there is a lot more that is needed, including if the underlying GenAI technology actually works,” he says. “You need great governance tools, but you also need great LLM models behind it.”

Jignesh adds: “Governance also spans the entire set of technologies that you have in your enterprise stack and GenAI is getting embedded everywhere. Often even within the same enterprise there are multiple LLM models, including internally hosted open-source models. Does the AI governance cover all that? That is a difficult task, and early governance tools may need to mature to cover the full spectrum. But, we have to start somewhere.”

Is IBM the right company to bring in safeguards?

This brings us to one of the key questions: is IBM well placed to provide the governance toolkit for businesses and governments?

“Here, the hyperscalers like Google, Microsoft and Amazon may have a bigger advantage [than IBM] as they often have their own LLMs that they can host and on their own hardware,” says Jignesh.

“Most likely, it will be easier for them to make a comprehensive offering with governance built-in. But, for customers who are a big IBM shop, this may be a way.”

Vaclav adds: “IBM built its brand by being a reliable partner to mature organisations in regulated industries. As the old saying goes: ‘Nobody got ever fired for buying IBM’. Part of this partnership is providing tools to help with implementing and enforcing governance rules and policies.”

We’ll give the final word to Julien, who believes that IBM’s “long-standing reputation as a leader in the technology industry” — in addition to its hefty investment in AI platforms — give it some key advantages.

“The trust in IBM may vary among CEOs and CTOs,” Julien says, “but historically, IBM has been considered a trustworthy partner for enterprise-level solutions, which can weigh positively for its governance tools.”