CANNES — Can artificial intelligence predict the outcome of an election? Back in November, Havas did exactly that.

When the world was betting on a Hillary win, Havas deployed what, like large tech firms, it is now calling its “cognitive” technologies, on 15 million news articles, candidates’ speeches and a billion social posts by 10 million US voters.

“The platform was predicting a Trump win even two to three weeks before the election,” Havas global head of marketing innovation Jason Jercinovic tells Beet.TV in this video interview.

“We were mystified. What came out of that was how Donald was doing that – he was targeting voters that had been switched from voting for Obama (using) very sophisticated targeting and pushing material to them to cause them to think in a certain way.”

Jercinovic’s crystal ball partly relied on topic classification technology from IBM Watson’s suite of cognitive APIs. Havas deployed the findings in an election-night special show by the UK’s ITV News.

Now Havas is building the platform in to an offering its calling Eagle.Ai, for determining the outcomes and key issues of other elections.

But, for Jercinovic, Eagle.Ai’s findings have actually given cause for concern – about the technology now available to both politicians and advertisers alike. The practices of Cambridge Analytica have concerned some observers.

“You can predict the future, outcomes, even influence those outcomes and effect behavior,” he tells Beet.TV. “Thus, we need to act with a mature responsibility and speak from a point of ethics. Now is the time to do that … we need to protect ourselves from ourselves.”

So Havas is involved in bringing together ad agencies to develop guidelines around how to use new AI technologies. Media agency executives are now sharing ideas for how to conduct consumer profiling using artificial intelligence, as one agency urges the industry to adopt its code of conduct to avoid damaging privacy violations.

Jercinovic says the power of AI applied in advertising could be huge – and also destructive.

“Potential violation of trust could be damaging beyond belief,” he says. “Many companies have deployed a set of APIs which you can basically interrogate a data set. That allows you to pull these insights out that can be very personal, can be very intimate…

“(With) 400 or 500 Facebook posts (analyzed), I can effectively map (a person) to Myers Briggs or an OCEAN personality demographic and infer a lot of things. Therein lays the existential challenge – with that power to have those insights, you can know more about a consumer than potentially they know themselves.”

The application of running those APIs on available consumer signals is significant, for anyone trying to target the right consumers.

Jercinovic imagines: “It’s not hard to determine inferences of things like sexual preference, political affiliation, purchase intent etc. and thus the responsibility is critical now for us to look at ways we can protect them, make sure they see the value of these exchanges.”

For one thing, Europe’s forthcoming General Data Protection Regulation (GDPR) make this sort of profiling highly questionable without explicit consumer opt-in permission.

And Jercinovic reveals agency rivals are now discussing with each other how best to exploit the new opportunities whilst also ensuring consumers’ wishes are respected.

“We’ve been putting forward a code of conduct around this, a system of trust which is based around self-regulation, industry-wide,” he says. “We share ideas across many of the holding companies.”

This video is part of Beet.TV’s AI Series from Cannes Lions 2017, presented by The Weather Company, an IBM Business. For more from the series, please visit this page.