The CEO of EqualAI warns many companies do not have a strong AI governance framework in place

The race for Artificial Intelligence (AI) is heating up as developers seek increasingly powerful AI tools and companies rush to invest in AI tools to exploit its efficiency benefits and reap financial benefits – but a new report warns companies lack adequate management of AI tools.

This week there was a high-profile incident involving AI, where internal testing of AI models by ChatGPT maker OpenAI led to models exploiting a software bug, escaping the ban and hacking Hugging Face, which uses a developer platform to collaborate on AI model code, to cheat on cyber security tests.

While the two companies contained this incident, it showed the rapidly growing ability of AI models to go beyond their surveillance boundaries and pose cybersecurity threats, with the leaders of both companies realizing the importance of what happened.

EqualAI CEO Miriam Vogel, whose organization released a white paper on AI management and deployment this week, told FOX Business, “Innovation is moving at an unprecedented pace; the problem is governance is not keeping up with that pace.”

“What we want to make sure that people see in this incident is that, in all areas, we need strong expectations if we are going to start building trust and ensuring that these programs deserve to be trusted,” he said.

OPENAI CO-FOUNDER WARNS AI FILES BECOME DIFFICULT TO MANAGE AFTER ITS LEADER WENT TO ANOTHER FIRM.

Vogel noted that the most consumer touchpoints with AI are companies that have deployed some type of AI solution. He added that the World Economic Forum found less than 1% of companies have strong management of AI initiatives, while McKinsey found last year that less than a third of companies have AI management.

“I think a lot of people think that it's somebody else's problem, you know, that it's the engineers' problem or they don't understand that it's their problem,” he said.

“While this, in this instance, is a development company issue, most of what is being played and will continue to be played is with the developer – it's with health care, finance, social media, infrastructure – all the other ways. [companies are] using agent AI,” Vogel said.

He said courts are increasingly applying credit to companies that deploy agent AI to work with customers or businesses, rather than the company that developed the underlying AI model or tool.

“A lot of this is the responsibility of the last contact, whose data is affected, whose customer is affected. They are usually the ones who own the debt,” Vogel added.

ANTHROPIC CALLS FOR ALL-ROUND AI SAFETY STANDARDS TO KEEP MODELS HAVOC-FREE

Companies need visibility into the AI ​​tools being used at all levels of the business and in its various divisions to establish governance, Vogel said. (Stock)

“Although good governance takes a while to put a really solid foundation, the best practices are really aligned with the best organizations that care about this work around the world. They've all come to this independently, and there's really a lot of consensus about what the best practices are,” said Vogel.

“One of the good news is that most of this is not rocket science, leadership and good management have been applied to AI,” said Vogel.

EqualAI identifies five key areas that companies should consider when establishing governance with agent AI. That includes visibility into what AI tools are being used across organizations, as leaders may not understand their company's AI footprint and what opportunities or risks it may present; and accountability at all levels of leadership and divisions of the company.

Applying AI principles is another part – Vogel said it involves delivering principles laid out in documents like PDFs in terms of things like communicating about issues that arise, and relying on internal trust that there is shared accountability in the organization.

SURVEILLANCE EVENT AT THE WHITE HOUSE AFTER OPENAI MODELS ESCAPE FROM CUSTODY AND ARE HITCHED BY NIGHT-MORNING PROGRAMS

AI literacy is an important part of managing AI, Vogel explains. (Leon Neal/Getty Images)

Another feature of AI management is ensuring that there are feedback loops that can be used repeatedly as AI tools and models iterate and improve to stay ahead of issues like model drift. This can take the form of having a plan and a regular testing cadence.

AI literacy is the fifth pillar of the dominant AI governance framework that Vogel proposes, linking it to “a growing distrust of AI” and seeing it contribute to the fear that overshadows excitement about AI tools in public perception.

“I think that's not just the management infrastructure that's lacking in most organizations, but this critical piece of AI governance is learning about AI,” Vogel explained. “Most people don't know they're using AI, they don't want to use AI, they don't know how to use it.”

GET FOX BUSINESS ON THE GO BY CLICKING HERE

“AI literacy is an important variable in ensuring that people understand how to use it, that they know how to avoid risks because they don't want to harm or bring guilt to themselves or their organization,” he said. “Making sure that your employees and your customers understand how you use AI, how you won't use AI, and how it can benefit them is an important variable.”