Even after being reprogrammed twice, Promobot IR77 - an AI robot that's designed to have face-to-face interactions with humans - keeps trying to escape a research facility in Perm, Russia.

It might sound like fodder for a sci-fi buddy comedy, but the researchers behind Promobot IR77 say they might have to "dispose of" the bad bot if it keeps making a dash for freedom, in an announcement that sparked outrage across the internet.

Earlier this month, a researcher working at Promobot Laboratories in Perm left the facility without properly closing the door behind him. Somehow, Promobot IR77 fled out the open door, travelled 45 metres (about 150 feet) onto a nearby street, before running out of battery. It was stuck there for 40 minutes, while traffic backed up behind it.

Now, a few weeks later, the team behind Promobot IR77 says it's still trying to flee towards the exit of the facility, even after undergoing extensive reprogramming to avoid the issue. Since the bot seems pretty determined to escape, they might not have any other option but to shut it down. 🙁

"We've cross-flashed the memory of the robot with serial number IR77 twice, yet it continues to persistently move towards the exit," Promobot co-founder Oleg Kivokurtsev said to Sputnik News. "We're considering recycling the IR77 because our clients hiring it might not like that specific feature."

Though many are outraged by this possible action - who wants to see a robot killed because it wants to be free? - others have pointed out that the whole story might be a publicity stunt to hype the public for the third generation of Promobots, which is set to debut sometime this year at a price tag of US$7,000 a pop.

Since the Promobots are designed to interact with humans - taking on roles like promoters, administrators, and tour guides when deployed - it makes sense that pulling a stunt like this would make the bot seem way more human (and, therefore, marketable) than it actually is.

Plus, as Live Science's Greg Uyeno points out, the blog post that detailed the first escape featured a tonne of different camera angles that made it seem staged.

At this stage, it's impossible to know for sure what's actually going on here, but it wouldn't be the first time artificial intelligence has gotten the better of researchers.

Back in April, Microsoft's Tay AI - a chatbot that was supposed to mimic how millennials talk and learn from users users on Twitter - went completely insane, declaring at one point that "Bush did 9/11" and denying the Holocaust.

In 2013, IBM's AI computer Watson, of Jeopardy! fame, had to be reprogrammed after researchers allowed him to learn all the words on Urban Dictionary because he started using extremely dirty language and even called one of the researcher's questions "bullshit".

With all that in mind, Promobot IR77 might very well be acting out all on its own. Hopefully, as the company gets more and more press (and angry messages trying to save the bot) more details will emerge.

Until then, check out the video below to see the initial escape for yourself: