On the Bluesky social network, you may notice a lot of drive-by responses from accounts that rarely or never post — they just reply to other accounts.

The reply pattern starts with a phrase like “I respectfully disagree” and follows it with a fatuous objection. Another pattern is to start by agreeing, then pivot to trying to start a fight.

These responses are clearly from LLMs. Some Bluesky users have even gotten the bots to post a haiku, spill their prompt, or argue with another bot.

Denise Paolucci (rahaeli on Bluesky) cofounded blog site Dreamwidth and previously worked in trust and safety at LiveJournal. She has a ton of experience with every possible form of social media bad actor and regularly posts on Bluesky about trust and safety. Lately, she’s been writing about our bot friends.

Paolucci thinks this particular wave of “reply guy as a service” bots are test runs by a spambot maker, who hopes to rent the bots out to other bad actors as a service: “It’s common for spamming software to be pitched as social media management.”

…

Bluesky are on the case and blocking the bots. You should report the bots as “unwanted replies” and use the word “bot” when the offender is clearly a bot.