Older men in Arkansas might see a close-up photo of President Donald Trump pumping his fist in the air, along with a message asking them to donate $30 to his campaign for a Super Bowl commercial.

Middle-aged women in California might see a photo of Trump pointing to a crowd, with a plea for them to give “any amount” to the campaign.

Before Election Day, politicians across party lines are expected to spend more than $1 billion to pelt voters with millions of these cheap online ads, which can be tailored to a voter’s most personal details — down to one household or even a single individual.

Experts warn that this ad-targeting system is still vulnerable to manipulation by foreign governments and domestic actors trying to influence the election, just as they did in 2016. Those attempts could become more sophisticated this year as tech companies wrestle with a dysfunctional federal election watchdog agency and deploy haphazard safeguards that still offer plenty of loopholes.

“There’s now so much money and attention spent online with so few rules that if you wanted chaos, that’s the place to go for chaos,” said David Karpf, a political communications professor at The George Washington University. “And there’s a bunch of people who want chaos.”

According to Facebook, Russia-connected accounts spent about $100,000 on Facebook ads during the 2016 presidential election. The ads seemed to fan division on polarizing issues such as gun control and race relations. That’s a fraction of the cost of a single 30-second spot on a major TV network.

But it was enough to stir up trouble. In response, Google, Facebook and Twitter instituted verification policies that require advertisers to confirm their identity using their organization’s tax identification number or other government ID. Twitter later banned all political ads.