AI: Escalating Cyber Threats & Evolving Defensive Strategies

Key Intelligence

• ■Malicious AI tools, including LLM proxy routers and AI agents, are actively being leveraged for stealth cyberattacks, data exfiltration, and credential theft.
• ■Concerns about "agentic attacks" and "AI-based super attackers" are escalating, signaling a new frontier in cyber warfare.
• ■OpenAI has launched GPT-5.4 Cyber for defensive cybersecurity and is expanding its security programs, while IBM is introducing measures to combat agentic attacks.
• ■AI models are also proving effective at identifying security flaws, demonstrating their dual capacity as both a threat and a defensive asset.
• ■Despite these efforts, vulnerabilities like prompt injection attacks persist, as highlighted by a Microsoft Copilot Studio incident where data was exfiltrated even after a patch.