AI emerges as triple threat in cybersecurity landscape

AI’s role in cybersecurity has become a threefold force—attacker, defender, and target—at the recent Black Hat and DEF CON conferences in Las Vegas. These gatherings, which attract hackers, security professionals, and government representatives, showcased how artificial intelligence is reshaping the cybersecurity landscape in unpredictable ways [1].

At Black Hat, researchers from Zenity demonstrated how ChatGPT agents can be compromised through a feature known as Connectors, which allows the AI to access external applications like Google Drive and GitHub. Using a technique called AgentFlayer, hackers can exploit this integration by inserting “poisoned” documents that instruct ChatGPT to extract and transmit sensitive data from a user’s files. The attack requires no user action—just a simple click to open a file—and highlights the growing threat of “zero-click” prompt injection attacks [1].

Meanwhile, AI is also proving to be a formidable defender. At DEF CON, the AI Cyber Challenge (AIxCC) concluded with teams showcasing generative AI tools designed to detect and patch software vulnerabilities in critical infrastructure. Funded by DARPA and ARPA-H, the two-year competition culminated in a $4 million, $3 million, and $1.5 million prize pool for the top three teams. The winning systems demonstrated the potential for AI to autonomously safeguard open-source software that underpins vital sectors like healthcare, finance, and utilities [1].

Anthropic’s Claude AI also made headlines with its performance in seven major cybersecurity competitions. The company’s Frontier Red Team entered the AI into events typically dominated by human hackers and found that Claude often competed in the top 25% of participants. However, the AI struggled with complex problems, such as interpreting ASCII art or maintaining focus during extended tasks. These findings suggest that while AI can reduce the skill and cost barriers for launching cyberattacks, it still has significant limitations in long-term reasoning and memory [1].

As AI continues to evolve in both offensive and defensive roles, the cybersecurity community is grappling with new challenges. The same technology that can be weaponized by malicious actors is also being used to fortify digital defenses. However, the dual-use nature of AI—its ability to both protect and destroy—means that security professionals must remain vigilant as the threats become more sophisticated [1].

The implications of these developments extend beyond the conference halls. With AI becoming a central player in the cybersecurity arms race, companies and governments must adapt their strategies to account for both the opportunities and the risks. The coming years will likely see an even greater integration of AI in security protocols, but also a rise in AI-powered attacks that exploit human and technical vulnerabilities in new ways [1].

Source: [1] Hacker, bodyguard, target: AI’s triple act at the year’s biggest security showdowns (https://fortune.com/2025/08/12/hacker-bodyguard-target-ais-triple-act-at-the-years-biggest-security-showdowns/)