Chess has already been plagued by AI. There are engines that can consistently beat, or tie with the best chess players in the world. The most famous instance of which is when Magnus Carlsen - the current best chess player to ever exist (I'm afraid that's not an overstatement) admitted that he wouldn't dream to ever be able to compete with the best chess engines.
These engines are usually taught using machine learning - this isn't good for
US poker players. If someone had the computing power required to run AI poker engines against less experienced players (likely by utilizing free-roller tournaments), then they would eventually create an AI that's substantially +EV in low stakes poker. The same process could in theory, be repeated for mid, or high stakes poker. However, this posses a new issue: you often have to invest a massive amount of money on the line in order to play against enough players. The engines act like players, it learns from the
hands it plays, the only difference is that the engine can play hundreds of hands a second. The only draw back is that the engine's learn substantially less from each hand it plays relative to human players. The engines will require millions of examples - or in this case, hands, in order to fully grasp all aspects of the game - this includes the "the exploitative element" of poker. the AI's will be trained to handle every possible situations regardless of how it's opponent reacts. And since the engines require so many hands, and learn so slowly relative to it's human counterparts, the engine will cost millions of dollars in buy-ins alone, and likely, years to be able to play enough hands against competent players.
I'm not saying that a highly skilled poker AI is impossible, but it would be extremely costly to produce and, like many other fabricated poker AI's that already exist, it would likely play 'robotically', it would be predictable. It would only ever make the perfect bet sizes, and the perfect decisions that would be most likely to benefit the engine. That's likely where the
poker site would have to step in and ban the player, for "playing too perfectly". I do believe that despite the drawbacks of engines, they will be an issue in the future, where lower and mid stakes players are slowly replaced by AI's that compete for the food scraps left behind by the what few human players still play. I do, however, think that high stakes poker ill be safe from the engines, the cost would be monumental, and the poker engines would hardly be +Ev in such competitive fields.
The resolution, ultimately, will have to come from the poker site's themselves, they'll require anti-cheat detection. the sites will need to own their own engines that know exactly what the best move is, and compare it's move to the players in order to determine whether or not it's a player - this is a heavily flawed system, but maybe, it'll be effective in the short term. I'm uncertain whether or not this is the 'only' solution, but, in the near future, these sites will need to develop new solutions to combat poker engines or else
online poker will be devoid of human players, and shortly after, shut down.
In summary, yes poker will soon be infected with poker engines (if it hasn't already been). The engines, while imperfect, will be able to be +Ev in any stakes that they are trained in - given the fat that they are trained with enough hands. And our only hope to stop them comes from the
poker sites developing the means to detect and ban the engines before they cause too much damage to those who play the game fairly, without the use the an engine.