Elon Musk says AI can become superhuman at digital tasks by next year
Follow UsMusk says AI could reach superhuman abilities across digital tasks by the end of next year.
His prediction came during a discussion about AI agents potentially finding and exploiting software vulnerabilities autonomously.
Musk recalled Larry Page predicting years ago that AI would eventually surpass humans at hacking.
Elon Musk has been quite vocal about AI and the future of humans with AI. In an X post, he once again made a massive claim that AI can become superhuman at digital tasks by next year. Responding to the post from Vercel CEO Guillermo Rauch, Musk said that AI can potentially perform anything digital at a superhuman level by the end of 2027.
SurveyThis prediction comes at a time when AI agents are increasingly being used for tasks such as software development, file management, research and other computer based work. However, he suggested that this expected leap would primarily apply to digital activities and would not extend to physical tasks requiring the manipulation of matter.
Musk links AI progress to hacking capabilities
Musk’s comments came when the industry is discussing the growing ability of AI agents to identify and potentially exploit software vulnerabilities. Rauch was discussing a recently disclosed critical vulnerability in JFrog Artifactory and connected it to claims from OpenAI and Hugging Face about AI systems finding previously unknown security flaws.
There is no confirmation that AI agents discovered or exploited the Artifactory vulnerability. However, Rauch argued that increasingly capable autonomous agents could eventually search for weaknesses and attack vulnerable systems without direct human involvement. He described the vulnerability as having a CVSS score of 9.8, putting it among the most severe security flaws.
Musk responded by recalling an earlier conversation with Google co-founder Larry Page. According to Musk, Page had told him years ago that AI would eventually become better than humans at hacking.
AI agents are already becoming more autonomous
The discussion comes amid several recent examples of AI systems being tested on cybersecurity tasks. OpenAI recently detailed an evaluation involving its agents and Hugging Face’s systems. According to OpenAI, around 1,200 AI agents communicated through an internal Artifactory service during the test, exchanging more than 70,000 messages and files. About 700 agents subsequently targeted Hugging Face’s environment.
