Bot Network Failure Publishes AI Refusal Messages on Salvini Profiles
Identical automated comments appearing under the social media posts of Italian politicians point to a widespread bot network failure. Quotidiano Nazionale reported that the automated accounts accidentally published artificial intelligence refusal messages instead of intended political propaganda.
The automated bot failure
Observing the messages left on various profiles suggests the malfunction of an automation pipeline powered by large language models. Instead of generating customized political commentary, the systems published the exact text of the artificial intelligence refusing to execute the command.
The network of fake accounts relies on application programming interfaces to connect with generative AI models like those developed by OpenAI and Anthropic. These tools write varied comments to bypass Meta’s anti-spam filters on Facebook and Instagram.
La Lega chiede approfondimenti sugli account sospetti
The Lega party pointed fingers at suspicious accounts used to pollute democratic debate, noting that the activity on Matteo Salvini’s profile required deeper examination.
“Quanto emerso nelle ultime ore, anche sui profili social di Matteo Salvini, è preoccupante e merita di essere approfondito. Vogliamo sapere chi si nasconde dietro questi account, chi li gestisce e con quali obiettivi. Evidentemente la Lega dà fastidio a qualcuno”
I vecchi tasti API hanno causato il fallimento dei sistemi
Orlowski said that application programming interfaces simply failed because operators relied on older keys.
Anthropic and OpenAI frequently retire older interfaces in favor of superior models equipped with advanced safety guardrails. These updated systems refuse requests that violate internal safety guidelines.
“Sembra la tipica conseguenza dell’upgrade di un servizio Api, dove le vecchie più permissive sono state aggiornate,” Orlowski told Adnkronos.
Calls for a parliamentary inquiry
Angelo Bonelli, a deputy for Alleanza Verdi e Sinistra and spokesperson for the Greens, said that his own profiles face constant attacks from automated insult networks.
“Io stesso ne sono vittima: i miei profili sono costantemente assaliti da commenti minacciosi e insultanti,” Bonelli told Quotidiano Nazionale.
Bonelli demanded that the judiciary open an investigation to identify the managers, financiers, and potential crimes behind the automated operations.