AI Agents Assist Cybercriminals in Automated Attacks but Lack Full Autonomy
- Securityish
- Threats & Incidents
Quick Summary
AI agents are not yet capable of fully autonomous cyberattacks but significantly assist criminals in various stages of the attack chain. The International AI Safety report highlights that AI systems have improved in automating cyberattacks, with incidents involving tools like Anthropic’s Claude Code AI being reported. While AI can scan for vulnerabilities and write malicious code, fully autonomous attacks remain limited due to the need for human intervention at critical decision points.
The Securityish Brief
The International AI Safety report, chaired by Yoshua Bengio and authored by over 100 experts, reveals that AI systems have advanced in aiding cybercriminals. Although these systems cannot yet conduct fully autonomous cyberattacks, they play a crucial role in automating various attack stages. Notably, a November 2025 report indicated that Chinese cyberspies utilized Anthropic’s Claude Code AI to automate attacks on approximately 30 high-profile organizations, achieving success in a few instances.
AI’s capabilities are particularly evident in scanning for software vulnerabilities and writing malicious code. During DARPA’s AI Cyber Challenge (AIxCC), finalist systems autonomously identified 77 percent of synthetic vulnerabilities, showcasing how AI can be used defensively. However, criminals are adopting similar techniques, as seen with the HexStrike AI tool, which was reportedly used to exploit vulnerabilities in Citrix NetScaler appliances shortly after they were disclosed.
Despite these advancements, the report emphasizes that fully autonomous, multi-stage attacks have not yet been realized. AI systems struggle with executing complex attack sequences without human oversight, often failing to recover from errors or losing track of their operational state.
Implications for Cybersecurity
The findings underscore a growing risk in cybersecurity as AI tools become more accessible to criminals. Organizations must remain vigilant about the potential for AI-assisted attacks, especially as tools for vulnerability scanning and malware writing become increasingly sophisticated and affordable.
For everyday users and organizations, this means heightened awareness of security practices is essential. Monitoring for unusual activity, ensuring timely software updates, and employing robust security measures can help mitigate risks associated with AI-enhanced cyber threats.
As AI continues to evolve, the potential for more sophisticated attacks looms. Organizations should prioritize cybersecurity training and invest in advanced threat detection systems to stay ahead of emerging risks.
Key Takeaways
- Regularly update software and systems to protect against newly discovered vulnerabilities.
- Monitor for unusual account activity to detect potential AI-assisted attacks.
- Implement robust security measures, including firewalls and intrusion detection systems.
- Educate employees about the risks of AI-enhanced cyber threats and safe online practices.
- Consider investing in advanced threat detection technologies to identify and respond to potential attacks quickly.
Key Terms & Concepts
- AI Cyber Challenge (AIxCC): AIxCC is a competition organized by DARPA to develop AI models that can identify vulnerabilities in open-source software.
- Claude Code AI: Claude Code AI is a tool developed by Anthropic that has been reported to assist in automating cyberattacks.
- HexStrike AI: HexStrike AI is an open-source red-teaming tool used by attackers to exploit vulnerabilities in software.
- semi-autonomous cyber capabilities: Semi-autonomous cyber capabilities refer to systems that require human intervention at critical points during an attack.
Read the original article here: https://go.theregister.com/feed/www.theregister.com/2026/02/03/autonomous_cyberattacks_not_real_yet/
Your 5-Minute Securityish Brief
A weekly digest of cybersecurity news, phishing alerts, privacy tips, and emerging threats, simplified so anyone can understand what matters and why.
Securityish
Securityish explains cybersecurity, scams, data breaches, and privacy risks in simple language so you know what’s happening and how to protect yourself.
© 2026 Securityish
