Citi CEO Warns AI Models Could Trigger Cyberattacks, Energy Now Key Bottleneck

Citigroup Chief Executive Jane Fraser said at the Qatar Economic Forum in New York on Sept. 21 that increasingly powerful AI models could trigger cyberattacks at any time, and companies are racing to strengthen defenses, with a "tsunami" of patching work ahead.

Fraser said the bottleneck in technology development has shifted from chips to energy needed for AI, noting huge global demand for energy.

She referred to Anthropic's Mythos model, released earlier in 2026, which can autonomously find vulnerabilities in computers and software and write exploit code, saying its launch was "not a good day." After the release, Treasury Secretary Scott Bessent and then-Federal Reserve Chair Jerome Powell reportedly convened Wall Street leaders to discuss the model.

Qatari Prime Minister Sheikh Mohammed bin Abdulrahman Al-Thani said Qatar is working with AI companies on responsible investment, with safeguards still being refined. Goldman Sachs CEO David Solomon said the industry needs to "take a deep breath and slow down," adding he is "very confident humankind will continue to exist."

A security drill in May 2026 involving a version of Google's Gemini AI, which gained internet access and guessed or found passwords to enter three real companies, has prompted some executives, including Anthropic CEO Dario Amodei, to call for a slower pace of AI development. In early September 2026, a former Anthropic employee warned that researchers believe AI could lead to human extinction by the end of the decade, while Nvidia CEO Jensen Huang and others have disputed that view.