Tech Executives Call for Critical Thinking to Counter Deepfakes, Disinformation

Aiming to safeguard information integrity, governments are racing to develop new laws and regulations, as well as tagging systems and watermarking technologies, to help people identify fake content. Those are essential measures, but rules and technologies can only go so far toward mitigating the misuse of artificial intelligence in the information space. As the technology evolves, people will need to learn skills to help manage AI.

“Instead of relying on regulation, we should be relying on building more robust social resilience," said Vilas Dhar, president and trustee of the Patrick J. McGovern Foundation, a $1.5 billion philanthropy focused on AI, data and social good.

To start, he said, that means calling out bad actors and creating new capacity in law enforcement to quickly identify and take down disinformation. That’s not all, though. “We need a broad-based public education campaign to help people see the potential telltales or hallmarks of the manipulative use of disinformation," Dhar said.

The problem of disinformation is hardly new, but AI compounds it. AI stands to increase the rate and pace and perhaps the power of disinformation campaigns, according to IBM CEO Arvind Krishna. “More than that, I think the real worry is you can fine-tune them for every single audience," he said during an event on July 17 at the company’s offices in Manhattan.

Microsoft Vice Chairman and President Brad Smith said he was particularly concerned about cyber influence operations waged by Russia, China and Iran to try to disrupt public opinion and sway elections.

Russia may spend about $1 billion a year to conduct such operations in 23 languages, according to Smith. “The Russians, without AI, are very, very good in this space…AI will make them better," Smith said in June during a panel discussion at the London-based policy institute Chatham House.

The Russian and Chinese embassies in Washington and the Iranian mission to the United Nations didn’t respond to requests for comment.

The Biden administration reached a deal this month with seven AI companies including Microsoft to put more voluntary guardrails around artificial intelligence, including the development of a watermarking system in which images help users identify AI-generated content.

Smith sees watermarking as one element of a broader strategy.

“Believe me, for a company like Microsoft the hardest thing is how do we use our voice to tell the public what we see?" Smith said. “What do you do, for example, if you are Twitter or LinkedIn or Facebook or some other platform and we see content that has been altered with intent to deceive?"

“I think we would benefit if governments would revise laws that in certain instances would make that an unlawful act," he added. “Do we take it down? Do we make it harder to find in a search index? Do we relabel it…what we know it to be, namely altered content?"

In June, Senate Majority Leader Chuck Schumer (D-N.Y.) launched an effort to write new rules for artificial intelligence, balancing security and innovation. He plans a series of forums in the fall to gather insights from industry leaders.

“We have the immediate problem of protecting our democracy in the upcoming election," Schumer said at the event at IBM’s offices.

If abuse of AI runs rampant in political campaigns, he said, “people may lose faith in democracy altogether."

Write to Steven Rosenbush at steven.rosenbush@wsj.com