AI-Powered Cyberattacks Growing Smarter, More Autonomous: Report

New Delhi: Artificial intelligence is rapidly transforming the cyber threat landscape, enabling hackers to automate sophisticated attacks and making it increasingly difficult for security experts to assess and counter emerging risks, according to a new report by Anthropic.

The AI company revealed that it banned more than 800 accounts linked to malicious cyber activities between March 2025 and March 2026. Its analysis indicates that cybercriminals are no longer using AI solely for planning and reconnaissance but are increasingly integrating it throughout the entire attack process.

The report found that around 67 per cent of the flagged accounts used AI for attack preparation tasks such as malware development and vulnerability research. However, researchers observed a notable shift towards more advanced operational activities after attackers gain access to targeted systems.

According to Anthropic, there is growing evidence that AI is enhancing the capabilities of cybercriminals, allowing them to conduct more sophisticated operations with greater efficiency.

The study also highlighted a rise in the number of medium- and high-risk threat actors using AI. Such actors accounted for 33 per cent of analysed cases during the first half of the review period, but their share increased to 56 per cent in the latter half.

While AI-assisted phishing campaigns have shown signs of decline, the use of AI in post-compromise activities has expanded significantly. These activities include identifying user accounts, exploring compromised networks, and moving laterally within affected systems to gain broader access.

Anthropic warned that cyberattacks are becoming increasingly autonomous as attackers adopt AI systems capable of linking multiple stages of an operation with minimal human involvement. These AI-driven tools can assist in planning, execution, and decision-making throughout an attack, reducing the need for constant oversight by human operators.

The report further noted that traditional methods of measuring attacker sophistication may no longer be sufficient. With AI helping less-experienced individuals perform technically advanced tasks, the gap between skilled and unskilled threat actors is narrowing.

Researchers also argued that widely used cybersecurity frameworks, including MITRE ATT&CK, are not fully equipped to address emerging AI-enabled threats. Existing models often fail to account for AI agents that can independently make tactical decisions, coordinate different phases of an attack, and execute actions autonomously.

As AI technology continues to evolve at a rapid pace, the report cautioned that organisations and security teams worldwide will face increasing challenges in detecting, understanding, and defending against next-generation cyber threats.
 

With inputs from IANS