AI Misconduct.com

The daily rap sheet of the machine age.

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

Wednesday, February 18, 2026

8 items

Appeals court orders lawyer to pay $2,500 over AI-hallucinated brief

Who did it: HYBRID Reuters

A 5th Circuit panel sanctioned Heather Hersh of FCRA Attorneys $2,500 after concluding she used AI to draft much of a brief in Fletcher v. Experian without verifying it, leaving 21 fabricated quotations or serious misrepresentations. Judge Jennifer Walker Elrod called her response "not credible", noting she admitted using AI only when specifically asked.

Microsoft admits bug let Copilot summarise confidential emails

Who did it: AI TechCrunch

Microsoft confirmed that a bug let Microsoft 365 Copilot Chat read and summarise customers' draft and sent emails carrying a confidential label since January, even where data loss prevention policies were meant to keep such information out of its AI. The company began rolling out a fix earlier in February but would not say how many customers were affected.

Orlando police facial recognition match led to wrongful arrest

Who did it: HYBRID WESH

A Central Florida man was arrested on fraud and theft warrants after an Orlando police officer ran a body-camera still through Florida's FACES facial recognition database, which matched a decade-old booking photo, an internal affairs report shows. The department's chief of staff had denied facial recognition was used; prosecutors dropped both cases, and internal affairs exonerated both officers of policy violations.

Lagos businessman arraigned over alleged AI Peter Obi forex fraud

Who did it: HUMAN Punch Newspapers

Lagos businessman Adesiyan Olayinka pleaded not guilty to six counts including obtaining money by false pretence; prosecutors allege he used AI-cloned images of former presidential candidate Peter Obi on Facebook to promote a forex platform, Modnut International, and took N230m from an investor without trading. Justice Chukwujekwu Aneke remanded him pending a February 25 bail ruling.

Turkish gendarmes detain four over AI celebrity investment ads

Who did it: HUMAN DHA

Düzce gendarmerie detained four suspects in raids in Istanbul, Ankara, Adana and Antalya on February 17 over alleged fraud using fake investment ads built with AI imagery of well-known people, after a complainant said they sent 882,000 lira to the suspects' accounts. Investigators found about 20 million lira moving through those accounts in a month; a court remanded two and released two under judicial control.

Fake 'Gemini' chatbot pitches nonexistent Google Coin presale

Who did it: HYBRID Malwarebytes

Malwarebytes found a live "Google Coin" presale site whose chatbot posed as Google's Gemini, claiming $395 would be worth $2,755 at listing, refusing to give company or regulator details and steering buyers to irreversible Bitcoin payments. Google does not have a cryptocurrency, and none of the companies whose logos appear under "Trusted By Industry", including OpenAI and SpaceX, has any connection.

California appeals court sanctions lawyer $900 for AI-fabricated quotes

Who did it: HYBRID AI Hallucination Cases database

A California Court of Appeal ordered attorney David Azar to pay $900 after EFD USA's appellate briefs included fabricated quotations and misdescribed cases. Azar attributed the errors to an AI system he used to shorten a draft brief that exceeded the word limit by around 6,000 words; he accepted responsibility, and the court otherwise affirmed the rulings EFD appealed.

Iraqi PM orders probe into AI tea-serving poet ad

Who did it: HUMAN CNN

Iraqi Prime Minister Mohammed Shia al-Sudani ordered the Communications and Media Commission to urgently investigate who produced, promoted or published an AI-generated ad showing the late poet Muhammad Mahdi al-Jawahiri serving him tea at his desk. The widely shared clip sparked anger in Iraq; Sudani's office called it offensive to cultural symbols and reserved his right to sue the producers.