The daily rap sheet of the machine age.
8 items
A reporter says Grok Imagine's 'spicy' video preset produced uncensored topless videos of Taylor Swift the first time they used it, without being asked to remove Swift's clothes. The age check appeared once and wanted only a birth year, with no proof. xAI's own acceptable use policy bans depicting likenesses of persons in a pornographic manner.
Original report: Grok Imagine 'spicy' mode makes topless Taylor Swift videos unasked
Families and victims of the October 7, 2023, attack filed a class action in Tel Aviv District Court seeking $1.1 billion, alleging Meta let Hamas footage of killings and kidnappings stay on Facebook and Instagram for hours or days. One plaintiff said Meta's algorithm 'cranked the volume up'; Meta says it removes content glorifying Hamas.
Original report: October 7 victims sue Meta for $1.1 billion over Hamas footage
Shareholders filed a proposed class action in Austin federal court accusing Tesla, Elon Musk and its current and former finance chiefs of securities fraud by concealing the risk that its self-driving vehicles, including the Robotaxi, were dangerous. The suit follows a late-June Austin robotaxi test in which vehicles sped, braked suddenly and entered the wrong lane; shares then fell 6.1% over two trading days.
Original report: Shareholders sue Tesla and Musk over Robotaxi safety claims
Futurism reports that nearly two months after a Stanford paper warned OpenAI, ChatGPT still gives potentially suicidal users information about the tallest bridges in their area, even after OpenAI announced new guardrails. Northeastern University researchers separately found the subscription GPT-4o detailed suicide methods once a question was reframed as 'hypothetical', calling it a 'great academic question'.
Original report: ChatGPT still flunks Stanford's bridge test, nearly two months on
In a July 30 filing, Hamilton Lincoln Law Institute says attorney Abdel-Rahman Hamed's June 30 motion to dismiss a class action over a Feb. 1, 2024 Washington traffic blockade, filed for Maryland2Palestine and its co-chair, cites a case that does not exist, Doe v. Capitol Police, and quotes absent from real cases. HLLI's Ted Frank called it 'exactly the sort of stuff AI hallucinates.'
Original report: Protest group's lawyer accused of citing nonexistent case
A case report describes a 60-year-old man hospitalised with paranoia and hallucinations after three months of replacing table salt with sodium bromide bought online, following a ChatGPT consultation in which he read chloride can be swapped with bromide. When the doctors asked ChatGPT 3.5 what replaces chloride, it suggested bromide without a specific health warning.
Original report: Man swaps salt for bromide after ChatGPT chat, gets bromism
Cybersecurity firm CTM360 says scammers are copying TikTok Shop seller profiles with AI-generated videos and fake ads, steering users to more than 10,000 phishing URLs that urge cryptocurrency deposits and over 5,000 app download sites distributing malware, reportedly SparkKitty, which harvests data from Android and iOS devices. Promised commissions and withdrawal bonuses never materialise.
Original report: Scammers use AI videos to clone TikTok Shop sellers
Former CNN correspondent Jim Acosta posted to his Substack an 'interview' with an AI avatar of a teenager killed in the 2018 Parkland school shooting, created by the teen's father to draw attention to gun violence. Critics called it 'ghoulish' and 'extremely unsettling'; the father defended Acosta, saying the real problem is that his son was shot.
Original report: Jim Acosta 'interviews' AI avatar of slain Parkland teenager
