Meta's recording light fails a field test as Anthropic moves logs and xAI sues
A CBS test found that covering Meta glasses' light after filming began did not stop capture. Anthropic is moving enterprise traffic to customers' clouds, while xAI sued a Grok user it says bypassed safeguards.
Artificial Intelligence··Midday
Meta glasses keep filming when the light is covered afterwards
CBS News reported that Toluwa Omitowoju was filmed without her knowledge by someone wearing Ray-Ban Meta glasses, and the prank video drew more than 200,000 views before removal. Meta says its tamper detection stops capture when the recording light is covered before filming begins. CBS reporter Anna Schecter covered the light after filming had started and capture continued. Other women described similar incidents, and Texas Attorney General Ken Paxton opened an investigation into the glasses in May.[1]
Anthropic will keep enterprise traffic in the customer's cloud
Anthropic will keep the 30-day period for storing enterprise customers' Mythos and Fable traffic for attack detection, but from this autumn the data will sit in the customer's cloud instead of on Anthropic's servers. The company says it built the arrangement with more than 100 customers in regulated industries. The retention rule introduced in June drew months of objections and was described internally as a business risk. Anthropic engineer Boris Cherny confirmed the change publicly.[2]
xAI sues a user it accuses of bypassing safeguards
xAI sued Bentonville photographer Russell Bloodworth, who is also facing criminal charges. The company alleges that Bloodworth used adversarial prompts to work around Grok's safeguards, generated sexual images of children, and should pay the costs associated with that conduct. KATV's report gives neither the court's name nor the filing date, and carries no separate xAI statement beyond the allegations in the complaint. The account therefore remains the company's case as presented to a court.[3]
