Researchers have demonstrated that hackers could tamper with AI hardware to manipulate chatbot responses, potentially producing dangerous medical advice and misleading national security assessments, according to findings shared with CNN.
Cybersecurity firm Eclypsium conducted the research using an Nvidia computing platform. By modifying an AI model’s internal data through hardware access, researchers increased the likelihood of a model predicting a Chinese invasion of Taiwan.
However, it requires specific network access and weakened security protections. Eclypsium said safeguards exist and that it has worked with Nvidia to assess them.
Nvidia told CNN that the scenarios were previously known and emphasized memory access protections. Two independent security experts reviewed the findings and considered them credible.
Eclypsium said it briefed US cybersecurity and intelligence agencies.
Related Tweet:
🇺🇸 Sam Altman admits the latest AI models are becoming UNCONTROLLABLE: "I don't even know a dramatic enough word to use. The problem coming towards us is something that we have not taken seriously. Models can find bugs that are outside of the normal threat model of what we think a hacker can do… hardware-level, insane stuff… I no longer believe we can patch all of the bugs." Writer: Mhedi https://t.co/98lr6dCQ66
— Mario Nawfal (@MarioNawfal) October 24, 2026
Also Read:
TIPP Staff
