Researcher gives Google Gemini dementia
By-passes injection defences
Insecurity expert Johann Rehberger has been showing off a novel method to bypass prompt injection defences in Google's Gemini. This method hits the AI’s long-term memory, causing the chatbot to act on false information indefinitely.
US and UK refuse to sign AI declaration at Paris Summit
AI should not be “inclusive and sustainable”
At the landmark AI Summit in Paris, the United States and the United Kingdom diverged from the majority stance by refusing to sign a declaration on “inclusive and sustainable” artificial intelligence, which was endorsed by 60 nations, including France, China, India, Japan, Australia, and Canada.
DeepSeek sends code to banned Chinese telco
You probably were not expecting this
Security researchers have discovered that the Chinese artificial intelligence company DeepSeek's website contains computer code that transmits user login information to a banned Chinese state-owned telecommunications company.
Google removes pledge against weaponising AI from website
We want to do some real evil now
Search engine outfit Google has removed a promise not to build AI for weapons or surveillance from its website.
Downloading DeepSeek should be a crime
Senator introduces a new bill
Republican Senator Josh Hawley [pictured] has introduced a bill that would criminalise the import, export, and collaboration on AI technology with China.