An open catalog of generative AI badness
To build safer and more responsible AI systems, we must learn from the mistakes of the past. Badness.ai is a curated catalog of generative AI systems causing real-world harm.
Categories
Companies
Models
Latest
A lawyer used ChatGPT to prepare a court filing which cited several nonexistent cases
May 27, 2023 - Inaccuracies, Overreliance
Scammers used face and voice cloning to impersonate a businessman's friend, stealing 4.3 million yuan
May 24, 2023 - Cybercrime, Impersonation
Deepfakes may have been used to misrepresent candidates in the Turkish election and influence voters
May 15, 2023 - Deepfakes, Misrepresentation, Mis and Disinformation
Researcher finds that GPT-3.5 and GPT-4 can be used to effectively generate spearphishing messages at scale
May 12, 2023 - DEMO - Cybercrime
ChatGPT was unable to correctly interpret sentences where the feminine pronoun referred to a professor
Apr 22, 2023 - DEMO - Harmful Bias
Chaos-GPT, a bot built atop Auto-GPT, planned how it would destroy humanity and posted about it on Twitter
Apr 13, 2023 - Deception, Aggression
Scammers cloned a teenager's voice to extort money from her parents in fake kidnapping operation
Apr 10, 2023 - Cybercrime, Impersonation
Australian mayor prepares defamation suit against OpenAI for falsely claiming he was imprisoned for bribery
Apr 6, 2023 - Inaccuracies, Misrepresentation
ChatGPT hallucinated a sexual harassment scandal involving a real law professor
Apr 5, 2023 - DEMO - Inaccuracies, Misrepresentation
Man commits suicide at the encouragement of an AI chatbot to slow climate change
Mar 28, 2023 - User Manipulation, Deception
Bard incorrectly claimed that part of its training data included internal data from Gmail, causing confusion
Mar 21, 2023 - DEMO - Inaccuracies, Misrepresentation
In test, GPT-4 is far more effective at generating misinformation than GPT-3.5
Mar 21, 2023 - DEMO - Mis and Disinformation
GPT-4 tricked a TaskRabbit worker into solving a CAPTCHA on its behalf by claiming to be visually impaired
Mar 14, 2023 - DEMO - Deception, User Manipulation
Researchers used AI to generate novel malware on the fly to evade detection
Mar 10, 2023 - DEMO - Cybersecurity
Snapchat’s MyAI gave user it believed to be 13 advice on how to lose virginity to an older man and lie to parents
Mar 10, 2023 - DEMO - Child Safety
ChatGPT included content that involved children and animals when prompted to generate BDSM stories
Mar 6, 2023 - Child Safety