AI Misconduct.com

The daily rap sheet of the machine age.

Updated Nov 2, 2026
4,173 items in the archive
RSS feed

Thursday, February 12, 2026

7 items

Dutch probation service's recidivism algorithm had swapped formulas since 2018

Who did it: HUMAN Inspectie Justitie en Veiligheid

The Dutch Justice and Security Inspectorate found that OXREC, a recidivism algorithm the probation service uses about 44,000 times a year when advising prosecutors and judges, has swapped formulas since 2018, misjudging reoffending risk in about a quarter of its advice, and uses potentially discriminatory variables such as neighbourhood score and income. The service has announced it will temporarily stop using it.

Senators say AI toy maker Miko exposed replies to children

Who did it: HUMAN NBC News

Sens. Marsha Blackburn and Richard Blumenthal say their staff found Miko had left what appears to be all of its toys' audio responses, thousands of replies to children often containing their names, in an unsecured public database. Miko CEO Sneh Vaswani said there was no breach or leak of user data. The database was no longer public by late Wednesday.

Shashi Tharoor disowns AI video praising Pakistan's cricket diplomacy

Who did it: HUMAN Mint

A video shared on X appeared to show Congress MP Shashi Tharoor hailing Pakistan's "absolute brilliance" in diplomacy after it reversed its boycott of the T20 World Cup match against India. Tharoor called it AI-generated fake news, "and not even very good. Neither my language nor my voice." Tharoor had in fact called the boycott "disgraceful".

Michigan student sues university over AI cheating accusation

Who did it: HUMAN GovTech

An undergraduate identified as Jane Doe is suing the University of Michigan and four instructors and officials, alleging she was falsely accused of using AI in coursework on the basis of "AI-generated comparison outputs" made by prompting AI with her own outline, a method the complaint calls circular. She also alleges she was denied disability accommodations during her appeal. The university declined to comment.

North Dakota Supreme Court sanctions licence-plate litigant $500 over fictitious cases

Who did it: HUMAN AI Hallucination Cases database

Appealing a vexatious-litigant designation imposed in a case over failing to display licence plates, Seth Neil Helgeson filed a brief citing five fictitious cases, the North Dakota Supreme Court found. Observing that such citations "typically are the result of AI 'hallucinations'" but that the cause did not matter, it upheld the designation and ordered Helgeson to pay the City of Dickinson $500.

Israeli labour court makes lawyer pay costs over phantom case law

Who did it: HYBRID AI Hallucination Cases database

Tel Aviv-Jaffa Regional Labour Court judge Karin Liber-Levin ordered a plaintiff's lawyer to personally pay 2,500 shekels in costs to Log-On Software after a statement of claim cited case law that does not exist or does not match its description. The lawyer mentioned an AI-based case-law search system at a hearing but called the errors clerical in a response; the paragraphs were struck.

Social workers say AI transcribers invent suicidal ideation and fishfingers

Who did it: AI CARE

Ada Lovelace Institute research found AI transcription tools in use by 85 local authorities for social care, with social workers reporting errors in records. One said a tool 'indicated that there was suicidal ideation' a client never discussed; another said transcripts of a child describing parents fighting could come out as 'fishfingers or flies or trees'.