AI Misconduct.com

The daily rap sheet of the machine age.

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

Wednesday, April 16, 2025

7 items

House report says DeepSeek bends up to 85% of answers to Party line

Who did it: HYBRID 台北時報

A US House Select Committee on the CCP report says up to 85 percent of DeepSeek's responses are altered or suppressed to suit CCP narratives, calls the chatbot a "profound threat" to US security, and alleges it channels US user data to the CCP via infrastructure tied to China Mobile. It deems unauthorized model distillation "highly likely".

House China committee opens probe into Nvidia chips reaching DeepSeek

Who did it: HUMAN The Straits Times

The House Select Committee on the Chinese Communist Party opened an investigation on April 16 into whether Nvidia knowingly provided DeepSeek with critical AI technology, potentially in violation of US rules. The committee says DeepSeek may have circumvented chip restrictions via intermediaries in Singapore. According to three people with knowledge, the Trump administration is debating barring Americans' access to DeepSeek's services.

House Democrats urge halt to any unapproved AI use by DOGE

Who did it: HUMAN Reuters

Nearly 50 House Democrats wrote to the White House budget office on Wednesday urging the Trump administration to stop using AI systems lacking formal approval such as FedRAMP, citing reported employee monitoring and DOGE's access to sensitive data. They asked whether Musk's xAI technology was in use and warned of conflicts of interest; the OMB did not immediately respond.

Telangana police serve IAS officer notice over AI Ghibli image

Who did it: HUMAN The South First

Telangana police served IAS officer Smita Sabharwal a notice under Section 179 of the BNSS for reposting, on 31 March, an AI-generated Ghibli-style image of bulldozers at the University of Hyderabad's Mushroom Rock amid protests over 400 acres of Kancha Gachibowli land. Officials allege such visuals were used to mislead the public; police confirmed the notice on 16 April.

Brazilian court denies AI-drafted habeas plea, refers lawyer to bar

Who did it: HYBRID AI Hallucination Cases database

Santa Catarina's Court of Justice on 16 April 2025 denied a habeas corpus petition it found drafted with AI by lawyer Micaelson Gelhen: of three Superior Court of Justice rulings cited, only one matched a real case, and the quoted passages did not remotely reflect what was decided. Calling it 'a complete disaster', judges sent the file to the bar's ethics tribunal.

Actors who licensed likenesses find AI clones pushing dubious cures, propaganda

Who did it: HUMAN The Star

South Korean actor Simon Lee told AFP his licensed AI avatar, sometimes posing as a gynaecologist or surgeon, promoted questionable health cures on TikTok and Instagram, and his contract barred him from getting the videos removed. British actor Connor Yeates found his Synthesia avatar promoting Burkina Faso's coup-installed president; Synthesia says some videos slipped its moderation.