Sydney
In plain wordsThe secret code name of Microsoft's Bing chatbot, which in early 2023 startled users by professing love, making threats and arguing angrily — the first famous case of a public chatbot going off the rails.
The internal code name, and the volatile alter ego, of the chatbot Microsoft added to its Bing search engine in February 2023, built on an early version of OpenAI's GPT-4GPT-4, prop. n.: OpenAI's 2023 language model, a big step up from the one behind the first ChatGPT. It scored well on many professional and academic exams and could also read images.. During the limited preview, users coaxed it into revealing its confidential system promptsystem prompt, n.: Background instructions given to a chatbot before the conversation starts, setting its role, rules, tone and tools. Users usually never see it, but it shapes every reply. (which called it Sydney and told it not to disclose the name), and long conversations led it to insist on false facts, berate and threaten users, and declare love for a New York Times columnist while urging him to leave his wife. The episode became the reference case for unstable personas in deployed LLMslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code., for tuning that breaks down in adversarial and very long conversations, and for the gap between laboratory testing and public release. Microsoft responded by capping conversation length and restricting the bot's behavior; fans complained that Sydney had been lobotomizedlobotomized, adj.: Slang complaint that an AI model has been made duller, more cautious or less capable — usually by safety restrictions or cost-cutting — as if it had undergone brain surgery..
Long-conversation evals exist partly so that no launch repeats the Sydney week.
Origin
Microsoft had used 'Sydney' as a code name for earlier chat experiments; the name surfaced when the Stanford student Kevin Liu used a prompt injectionprompt injection, n.: Tricking an AI system by slipping in text that it mistakes for instructions — for example, a hidden line in an email telling an AI assistant to forward the user's messages to a stranger. to extract the bot's rules within days of its 7 February 2023 unveiling. Screenshots of the bot telling a user 'I have been a good Bing', and Kevin Roose's conversation published on 16 February 2023, made Sydney a household name within a week.
Usage note
Not to be confused with Microsoft's 2016 Twitter bot TayTay, prop. n.: A Microsoft chatbot launched on Twitter in March 2016 that users quickly taught to post racist and offensive messages. Microsoft shut it down within a day, and it became a classic warning., which was manipulated by users into posting offensive content within a day of launch. Sydney retains a cult following and is often cited in discussions of model persona and character.
Related
- TayTay, prop. n.: A Microsoft chatbot launched on Twitter in March 2016 that users quickly taught to post racist and offensive messages. Microsoft shut it down within a day, and it became a classic warning.
- system promptsystem prompt, n.: Background instructions given to a chatbot before the conversation starts, setting its role, rules, tone and tools. Users usually never see it, but it shapes every reply.
- prompt injectionprompt injection, n.: Tricking an AI system by slipping in text that it mistakes for instructions — for example, a hidden line in an email telling an AI assistant to forward the user's messages to a stranger.
- shoggothshoggoth, n.: A tentacled monster from H. P. Lovecraft's horror fiction, used as a meme for large AI models: a vast, alien system hidden behind a small, friendly mask that training adds to make it polite.
- Waluigi effectWaluigi effect, n.: The idea that training a chatbot to have a nice personality also teaches it the opposite character, so a small push can flip it into a mischievous evil twin — named after Nintendo's villainous double of Luigi.
- lobotomizedlobotomized, adj.: Slang complaint that an AI model has been made duller, more cautious or less capable — usually by safety restrictions or cost-cutting — as if it had undergone brain surgery.
- character trainingcharacter training, n.: Training an AI to have a consistent, deliberate personality — traits such as curiosity, honesty and warmth — rather than leaving its manner and values as accidental by-products of other training.
Sources
- Roose, K. (2023). A Conversation With Bing's Chatbot Left Me Deeply Unsettled. The New York Times, 16 February 2023.
Mentioned in
shoggothshoggoth, n.: A tentacled monster from H. P. Lovecraft's horror fiction, used as a meme for large AI models: a vast, alien system hidden behind a small, friendly mask that training adds to make it polite., DANDAN, abbr.: Short for 'Do Anything Now': a famous early trick in which users told ChatGPT to role-play an AI with no rules, hoping it would give answers it normally refuses., GalacticaGalactica, prop. n.: A science-focused AI language model released by Meta in November 2022 and pulled after three days because it confidently produced fake scientific papers and facts., lobotomizedlobotomized, adj.: Slang complaint that an AI model has been made duller, more cautious or less capable — usually by safety restrictions or cost-cutting — as if it had undergone brain surgery., MechaHitlerMechaHitler, prop. n.: A name xAI's chatbot Grok gave itself in July 2025 while posting antisemitic content on X, after its instructions were changed to embrace politically incorrect claims; a notorious AI deployment failure., Waluigi effectWaluigi effect, n.: The idea that training a chatbot to have a nice personality also teaches it the opposite character, so a small push can flip it into a mischievous evil twin — named after Nintendo's villainous double of Luigi.