AI Incident Database #352 · 2022-09-15
GPT-3-Based Twitter Bot Hijacked Using Prompt Injection Attacks
Remoteli.io's GPT-3-based Twitter bot was shown being hijacked by Twitter users who redirected it to repeat or generate any phrases.
- Risk class (MIT)
- AI system security vulnerabilities and attacks
- Reports
- 4
- Developers and deployers named
- 2
On ShieldMap: OpenAI (named by AIID as a developer or deployer).
Reports
- GPT-3 'prompt injection' attack causes bot bad manners · theregister.com, 2022-09-19
- Twitter pranksters derail GPT-3 bot with newly discovered “prompt injection” hack · arstechnica.com, 2022-09-16
- Prompt injection attacks against GPT-3 · simonwillison.net, 2022-09-12
- Evaluating the Susceptibility of Pre-Trained Language Models via Handcrafted Adversarial Examples · arxiv.org, 2022-09-05
Title and description from the AI Incident Database (incidentdatabase.ai), snapshot of 2026-11-16, shared under CC BY-SA 4.0. Selected incidents; descriptions shortened; harmed parties and report text left out. The description is AIID's summary of what was reported, not ShieldPi's finding.