Microsoft accidentally revives Nazi AI chatbot Tay, then kills it again

Microsoft today accidentally re-activated "Tay," its Hitler-loving Twitter chatbot, only to be forced to kill her off for the second time in a week.

Tay "went on a spam tirade and then quickly fell silent again," TechCrunch reported this morning. "Most of the new messages from the millennial-mimicking character simply read 'you are too fast, please take a rest,'" according to the The Financial Times. "But other tweets included swear words and apparently apologetic phrases such as 'I blame it on the alcohol.'"

The new tirade reportedly began around 3 a.m. ET. Tay's account, with 95,100 tweets and 213,000 followers, is now marked private. "Tay remains offline while we make adjustments," Microsoft told several media outlets today. "As part of testing, she was inadvertently activated on Twitter for a brief period of time."

Microsoft designed Tay to be an artificial intelligence bot in the persona of a young adult on Twitter. But the company failed to prevent Tay from tweeting offensive things in response to real humans. Tay's first spell on Twitter lasted less than 24 hours before she "started tweeting abuse at people and went full neo-Nazi, declaring that 'Hitler was right I hate the jews,'" as we reported last week. Microsoft quickly turned her off.

Some of the problems came because of a "repeat after me" feature, in which Tay repeated anything people told her to repeat. But the problems went beyond that. When one person asked Tay, "is Ricky Gervais an atheist?" the bot responded, "ricky gervais learned totalitarianism from adolf hitler, the inventor of atheism."

Microsoft apologized in a blog post on Friday, saying that "Tay is now offline and we’ll look to bring Tay back only when we are confident we can better anticipate malicious intent that conflicts with our principles and values."