BENGALURU: Salil Parekh, chief executive of Infosys, pointed to emerging risks in artificial intelligence systems, referring specifically to vulnerabilities linked to Anthropic’s Mythos.
Parekh, during the Q4 earnings press conference said that current observations suggest AI systems may be more exposed than previously understood.
“It really… is exposing more vulnerabilities than one thought possible previously. However, other models are also exposing vulnerabilities.”
He did not provide technical details but indicated that as organisations adopt AI at scale, new risks are becoming visible, particularly around system behaviour, security, and reliability.
Parekh also described these developments as a potential opportunity for Infosys.
“My sense is it may also open up opportunities for Infosys… how can we make sure that you don’t succumb to that vulnerability.”
He said the company is already examining ways to help clients address such issues.
“We are looking at it… how can we make sure that your vulnerabilities are better and quickly protected.”
Anthropic’s Mythos is an advanced AI model with strong cybersecurity capabilities. It can identify and sometimes exploit previously unknown software vulnerabilities, including zero-day flaws. Because of these risks, access is tightly restricted to selected organisations.
Mythos is considered a dual-use system, useful for both defence and potential misuse, and highlights growing concerns about security, control, and reliability in powerful AI systems.