AI bots bypass social media safeguards on X, Meta and other platforms
Researchers demonstrated that AI bots can easily bypass policy enforcement mechanisms on major social media platforms including X, Meta, Reddit, and others, with high school-level interns capable of launching test bots using readily available technology. The bots were designed to spread misinformation, manipulate discussion, and commit fraud, and test accounts remained active after deployment.
What happened
Researchers at the University of Notre Dame tested bot policy enforcement on eight social media platforms: LinkedIn, Mastodon, Reddit, TikTok, X, and Meta platforms (Facebook, Instagram, and Threads). They successfully launched benign test bots on every platform.
Reddit, Mastodon, and X presented no challenge to bot deployment, with researchers describing enforcement as trivial. Meta platforms required multiple attempts and resulted in three suspensions before a test bot was successfully launched on the fourth attempt; TikTok presented modest challenges due to frequent CAPTCHAs.
High school-level interns with minimal training were able to launch test bots using readily available technology. As of the study's publication date, all test bot accounts and posts remained live on the platforms.
Reported impact
Certain phrases on X trigger bots to flood user accounts and inboxes. A researcher discovered this when searching for an artist, receiving 69 replies and dozens of message requests, many from automated accounts repeating phrases or offering unrelated services.
AI bots on social media are designed to manipulate public discussion, incite hate speech, spread misinformation, and enact fraud and scams. Researchers noted fake support accounts responding to posts containing keywords like 'Metamask wallet' and 'hacked'.
Index analysis
Platforms have published policies on using bots and created technical mechanisms to enforce those policies, but researchers found these mechanisms ineffective. The study concluded that social media platforms are not doing enough to stop harmful AI bots.
Governance controls
X Premium was introduced to address the bot problem, but researchers found the issue far from resolved. Some bots posing as real users have obtained verified accounts on the platform.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.