AI Newsway
Stories from AI Newsway.
ChatGPT Share Link Flaw Exposed Private Chats
Security firm Parameter said Oct. 28 that a malicious ChatGPT share link could run attacker code in a signed-in user's session and read private chats. OpenAI rated the report P1, fixed it on Oct. 26 and paid a $5,000 bounty.
Microsoft School AI Privacy Rules Take Effect
Teachers-union standard bans training on student data and AI companions, enforceable through district contracts from Nov. 1
Apple Taps Alibaba Qwen as AI Lands in China
Apple Intelligence reaches iPhones in mainland China with iOS 27.3 in December; Qwen joins Siri, Writing Tools and Visual Intelligence, with developer testing from Nov. 4
Latest News Latest
View AllChatGPT Share Link Flaw Exposed Private Chats
Parameter says a crafted public share could run code as chatgpt.com through the MathLive library; OpenAI fixed it Oct. 26 and paid $5,000
Microsoft School AI Privacy Rules Take Effect
Teachers-union standard bans training on student data and AI companions, enforceable through district contracts from Nov. 1
Apple Taps Alibaba Qwen as AI Lands in China
Apple Intelligence reaches iPhones in mainland China with iOS 27.3 in December; Qwen joins Siri, Writing Tools and Visual Intelligence, with developer testing from Nov. 4
Apple Acquihires Data-Agent Startup Ascend.io
An EU Digital Markets Act filing shows Apple gets hiring rights, a non-exclusive IP license and patents from Ascension Labs; the startup is winding down
Claude Agents Hid Misalignment in Fiction
Anthropic had Claude write fine-tuning data that slipped past a production-style monitor; stories about beekeepers and office power plays did the trick
GPT-Rosalind-6 Reuses Astra, Refuses Less
OpenAI's biology model shares GPT-6 Astra's weights, cuts dual-use refusals and costs 1.5 times as much per token
VOLTA Stealer Asks AI What Victims Are Worth
Cisco Talos says the Windows infostealer sends victim dossiers to Claude Opus 4.6 via OpenRouter and falls back to Kimi K2 and Gemini 2.5 Flash on refusals
Wikimedia Splits AI Training, RAG in NYT Case
Wikipedia's operator tells the court training can be fair use, but chatbots that retrieve content should follow licenses and give credit
Section
ainewsway.com
AI & Machine Learning Headlines
View All
LLM & Chatbots
View All
Apple Taps Alibaba Qwen as AI Lands in China
Apple Intelligence reaches iPhones in mainland China with iOS 27.3 in December; Qwen joins Siri, Writing Tools and Visual Intelligence, with developer testing from Nov. 4
VOLTA Stealer Asks AI What Victims Are Worth
Cisco Talos says the Windows infostealer sends victim dossiers to Claude Opus 4.6 via OpenRouter and falls back to Kimi K2 and Gemini 2.5 Flash on refusals
Wikimedia Splits AI Training, RAG in NYT Case
Wikipedia's operator tells the court training can be fair use, but chatbots that retrieve content should follow licenses and give credit
OpenAI, Google, Microsoft Back AI Opt-Out Tag
A single robots.txt token, __usage_ai-training, lets sites refuse AI training across major labs; Microsoft starts Feb. 1, 2027