The next phase of AI-driven cybersecurity attacks is expected to be unprecedented. This warning comes from Anthropic, an AI firm, which recently revealed that its new AI model, named Mythos, could exploit online security weaknesses much faster than previously seen.
Anthropic is not alone in these concerns. Earlier in December, OpenAI had similarly cautioned that its forthcoming models would present a high risk in cybersecurity. Experts believe that AI can not only escalate existing threats but also create new software vulnerabilities quickly.
The emergence of AI agents, software that can perform tasks without human input, raises the stakes. According to some experts, a single AI agent could identify and exploit vulnerabilities more efficiently than many human hackers combined.
This shift is being described as a significant turning point in cybersecurity by Shlomo Kramer, a leader in the field. Anthropic disclosed details about Mythos through a blog post, which was later leaked. The company blamed the leak on a mistake within its content management system.
Currently, Mythos is equipped with capabilities that surpass other models, indicating a wave of advanced tools that could far outpace security measures. To address this potential threat, Anthropic is allowing select organisations to test the model ahead of its full release.
The company is also alerting government officials about the risk of major cyberattacks facilitated by AI technologies. As new models from various companies continue to emerge, the challenges for cybersecurity will likely escalate, intensifying the ongoing battle between attackers and defenders.
Test Your Understanding
How much do you know?
