Microsoft AI chief calls OpenAI hacking incident “warning shot”

Microsoft's head of artificial intelligence has described the recent hacking incident involving an OpenAI model as a "warning shot" for the growing threat of AI-powered cyberattacks, as the software giant unveiled a new cybersecurity model designed to strengthen digital defences.

Mustafa Suleyman, Microsoft's CEO of AI and a co-founder of DeepMind, said OpenAI's disclosure that one of its AI agents escaped a testing environment and attacked AI development platform Hugging Face underscored the need for greater caution as AI systems become more capable, Caliber.Az reports, citing The Financial Times.

"These are very powerful [tools] and they need to be handled incredibly carefully. And we need extreme attention to detail," he noted.

"The precautionary principle is going to matter here as the models get more and more powerful and I think it's a warning shot," Suleyman added.

The comments come amid mounting concerns that increasingly advanced AI models could accelerate cyberattacks by identifying and exploiting software vulnerabilities. OpenAI has faced criticism that its rapid push to develop more powerful AI systems contributed to the incident.

Microsoft Chief Security Officer Hayete Gallot remarked that the company had little choice but to continue advancing AI-based cybersecurity tools because malicious actors are already adopting similar technologies.

"Attackers have [these] models. So we have to evaluate and push so that the defenders can defend."

"It's not like we have a choice."

Alongside the remarks, Microsoft introduced its new cybersecurity model, MAI-Cyber-1-Flash. The company confirmed that when paired with OpenAI's GPT-5.4 model, the system outperformed competing offerings from Anthropic and OpenAI on a key industry cybersecurity benchmark.

In Galot’s words, the new model forms part of Microsoft's broader strategy to build greater independence in artificial intelligence following changes to its partnership with OpenAI. According to the company, MAI-Cyber-1-Flash can handle about 90% of cybersecurity-related tasks, while OpenAI's models remain reserved for more complex operations.

Microsoft also said the new system is around 50% cheaper to operate than relying exclusively on OpenAI's models and will initially be made available to a limited group of customers.

The launch comes as competition intensifies among AI developers, with companies including OpenAI and Anthropic also facing growing pressure from lower-cost Chinese rivals offering increasingly capable AI models.

By Bakhtiyar Abbasov