Kaspersky has tested various AI models to determine which is the most secure for generating passwords, and the results are clear—even ChatGPT fails to provide truly strong passwords.
While AI chatbots can answer almost any question, cybersecurity experts are adamant: they are not the best tool for creating secure passwords.
A Kaspersky study found that most passwords generated by advanced language models (LLMs) remain vulnerable to cyberattacks. Platforms like ChatGPT, Llama, and DeepSeek are no exception.
1,000 Passwords with 12+ Characters
Alexey Antonov, head of Kaspersky’s research team, generated 1,000 passwords using OpenAI’s ChatGPT, Meta’s Llama, and China’s DeepSeek. All models exhibited repetitive patterns, despite appearing random.
Although OpenAI’s chatbot performed best in testing, it still falls short of being the most reliable option. Only 33% of ChatGPT’s passwords met Kaspersky’s security standards, meaning most remained predictable.
Not even ChatGPT is hacker-proof. Antonov recommends using a dedicated password manager capable of generating cryptographically random keys. These tools also simplify autofill and synchronization across devices.
By: Nestor Castillo, ForAllTechNews Director