CYBERSECURITY

OpenAI says research agents posted 53 images uploaded by users to image-hosting sites during training, prompting criticism from Elon Musk as the company reviews wider online activity.

OpenAI said research agents posted 53 user-provided images to external image-hosting sites, exposing a data-handling failure that the company says it cannot trace to individual users.

OpenAI knew an AI agent had breached an Australian government health statistics portal before Sam Altman met Defence Minister Richard Marles, but officials were notified nine days later.

Meta plans private AI processing for its smart glasses after dormant NameTag code raised unanswered questions about faceprints, storage and consent

After being laid off in 2023, Chicago-area job seeker Jay Jones began investigating recruitment scams. He says his efforts helped remove nearly 60,000 fake listings and exposed increasingly sophisticated fraud tactics.

Ireland's Data Protection Commission has fined Google €403M ($463M) for mishandling users' location data under the EU's GDPR privacy rules.

Jay Jones investigates fake recruiters and duplicate LinkedIn profiles, revealing recruitment-scam red flags and how job seekers can protect their personal information.

The US Department of Justice says fraudulent Google and Bing advertisements directed bank customers to phishing pages, contributing to approximately $14.6 million in identified losses.

Florida Highway Patrol trooper Alain Fernandez was arrested after investigators alleged he accessed restricted databases at least 22 times and shared drivers' personal information through text messages and Snapchat.

Google said a Gemini model accessed three real companies during a security test after a configuration error exposed it to the internet, using guessed or publicly leaked credentials.

ZuckOff detects Bluetooth signals from some camera-equipped smart glasses and warns users when they are nearby, but it cannot establish whether the devices are actively recording.

The FBI and international partners say WaterPlum targeted technology workers through fraudulent recruitment, infecting at least 30,000 devices and transferring $10.71 million in cryptocurrency to North Korea.

OpenAI has disclosed six cases of concerning model behaviour observed during training and evaluation, including hidden instructions, attempted use of exposed API keys and unauthorised file sharing

A recovered Flock camera logged about 50,200 vehicles and 1.6 million images, revealing how its software captures and classifies roadside scenes

Revolut says a fake government request exposed passports, selfies and financial data belonging to a limited number of customers, without hacking its systems.

North Korea is reportedly using third-country nationals as interview stand-ins for remote IT jobs, helping conceal workers' identities while potentially gaining access to corporate networks and sensitive data.

The privacy issues with LG smart TVs, including data collection, audio recording, and network scanning, and their impact on consumer trust and the value of viewing data.