Mindgard Finds Sora 2 Vulnerability Leaking Hidden System Prompt via Audio
AI security firm Mindgard discovered a flaw in OpenAI’s Sora 2 model, forcing the video generator to leak…
OpenAI
ChatGPT violated copyright laws, German court rules
OpenAI ordered to pay undisclosed damages for training its language models on pop songs without permissionA court in Munich has ruled that OpenAI’s chatbot ChatGPT violated German copyright laws by using hits from top-selling musicians to train its lan…
SoftBank sells stake in Nvidia for $5.8bn as it doubles down on OpenAI bets
Shares in chipmaker Nvidia fall after announcement by Japanese investor, with other tech shares also slippingBusiness live – latest updatesThe Japanese technology investor SoftBank intensified the debate about valuations in the artificial intelligence …
New ChatGPT Vulnerabilities Let Hackers Steal Data, Hijack Memory
Seven vulnerabilities in ChatGPT (including GPT-5) allow attackers to use ‘0-click’ and ‘memory injection’ to bypass safety features and persistently steal private user data and chat history. Tenable Research exposes the flaws.
SesameOp Backdoor Abused OpenAI Assistants API for Remote Access
Microsoft researchers found the SesameOp backdoor using OpenAI’s Assistants API for remote access, data theft, and command communication.
SesameOp Malware Abuses OpenAI API
A component of the newly discovered SesameOp backdoor uses the API to store and relay commands from the C&C server.
The post SesameOp Malware Abuses OpenAI API appeared first on SecurityWeek.
Who is Zico Kolter? A Professor Leads OpenAI Safety Panel With Power to Halt Unsafe AI Releases
Kolter leads a panel at OpenAI that has the authority to halt the ChatGPT maker’s release of new AI systems if it finds them unsafe.
The post Who is Zico Kolter? A Professor Leads OpenAI Safety Panel With Power to Halt Unsafe AI Releases appeared firs…
OpenAI Introduces Aardvark, an AI Security Agent Powered by GPT-5
OpenAI has announced the launch of Aardvark, an autonomous AI security agent powered by GPT-5 that aims to revolutionize how organizations discover and fix software vulnerabilities. The new tool, currently available in private beta, represents a signif…
Has OpenAI really made ChatGPT better for users with mental health problems?
Prompts indicating suicidal ideation got alarming replies, which experts say shows ‘how easy it is to break the model’An OpenAI statement released this week claimed the company had made its popular service ChatGPT better at supporting users experiencin…
‘ChatGPT Tainted Memories’ Exploit Enables Command Injection in Atlas Browser
LayerX Security found a flaw in OpenAI’s ChatGPT Atlas browser that lets attackers inject commands into its memory, posing major security and phishing risks.