Sophos predicts that artificial intelligence (AI) will face growing threats from vulnerabilities and malware in the coming year. The cybersecurity solutions provider has outlined this expectation as part of its 2025 predictions, emphasizing the increasing risks associated with large language models (LLMs) like ChatGPT.
“As vulnerabilities and malware targeting AI tools emerge, attackers are finding ways to exploit them,” said Christopher Budd, director of Sophos X-Ops. “Over the next year, we’ll see a clearer picture of these risks. Users and security professionals must work together to address these vulnerabilities and safeguard against the inevitable attacks that follow.”
Generative AI has also lowered barriers for cybercriminals. Opportunistic attackers are using AI platforms to generate phishing lures and even recreate ransomware code, explained Aaron Bugal, field CTO at Sophos. While these AI-generated threats often lack sophistication, their sheer volume creates noise, making it harder to identify serious risks.
Sophos also highlights the growing complexity of AI applications. Ben Gelman, senior data scientist at Sophos, predicts an increase in multi-agent systems where multiple LLMs work together to complete more advanced tasks. He noted that this approach could be used in areas such as cybersecurity, automated penetration testing, and customer service systems.
“Future advancements in AI will likely be incremental,” Gelman said. “We’ll focus on optimizing the current capabilities of LLMs, improving power efficiency, and reducing costs.”
With AI deeply embedded in operations, Sophos underscores the importance of proactive measures to mitigate its vulnerabilities. As AI tools continue to evolve, so do the strategies of those looking to exploit them.