That’s one of a number of ongoing efforts to compromise the language bot, threats that could lead to highly targeted phishing attacks.