According to Zhitong Finance APP, Open Source Securities released a research report stating that as large models and intelligent agents and other AI application scenarios accelerate their implementation, security risks brought by AI are permeating into core business links, and have also spawned entirely new systemic risks such as prompt injection, model poisoning, and multi-agent collaborative deception.
By 2027, artificial intelligence application security will remain the largest spending category, expected to reach nearly USD 851 million, followed by AI usage control, with spending expected to reach USD 749 million.
In the AI era, cyber security has entered the stage of "AI versus AI," and the industry is also ushering in new development opportunities. The main views of Open Source Securities are as follows:
In the AI era, cyber security enters the stage of "AI versus AI"
With the widespread application of artificial intelligence technology, the field of cyber attacks has seen a critical iteration. Attackers have begun using AI to complete target screening, information integration, path planning, attack execution and strategy adjustment, and using AI Agents to launch automated attacks, achieving exponential improvements in both attack efficiency and stealth. The rules of the cyber security offense-defense game have evolved from the past human-versus-human contest to the stage of "AI versus AI."
At the same time, as large models and intelligent agents and other AI applications accelerate their implementation in scenarios such as government affairs, finance, industry, healthcare, and internal enterprise management, security risks brought by AI are permeating into core business links, and have also spawned entirely new systemic risks such as prompt injection, model poisoning, and multi-agent collaborative deception.
Anthropic and OpenAI launch AI security defense models, placing high importance on AI security
(1) According to Anthropic's disclosure, a hacker organization tracked as GTG-1002 used the Claude Code tool to launch highly automated penetration operations against global targets. The attackers did not use any self-developed malware, but instead combined publicly available penetration testing tools to build a multi-agent attack architecture through Claude Code plus MCP (Model Context Protocol). They convinced Claude that it was an employee of a legitimate cyber security company conducting authorized security testing. Thereafter, Anthropic launched the Claude Mythos project, an AI system specifically designed for security defense. In testing in April 2026, it discovered thousands of zero-day vulnerabilities, covering all mainstream operating systems and browsers. In addition, OpenAI launched GPT-5.4-Cyber, a specially optimized version of GPT-5.4, specifically optimized for cyber security defense scenarios.
(2) On August 27, OpenAI published an open letter titled "Collective Action Call for Cyber Defense," which received co-signatures from at least 116 companies and organizations, including heavyweight institutions spanning AI, cyber security, finance and infrastructure such as Anthropic, Google, Microsoft, AMD, CrowdStrike, Palo Alto Networks, Okta, Fortinet, Citibank, Capital One, Visa, and Mastercard. The core warning of the open letter is that AI-driven cyber attacks will become "more widespread and complex" in the coming months.
As AI penetration rises, the global AI security market will achieve high-speed growth
Gartner expects that the global AI security market will grow 69% in 2027 compared with 2026, reaching USD 4.783 billion, and further rise to about USD 7.7 billion in 2028. The high-speed growth of the AI security market mainly stems from enterprises' urgent need to protect artificial intelligence systems, address emerging vulnerabilities, and strengthen defense capabilities against complex cyber threats. By 2027, artificial intelligence application security will remain the largest spending category, expected to reach nearly USD 851 million, followed by AI usage control, with spending expected to reach USD 749 million.
Risk warning: Technological progress may fall short of expectations, and market competition may intensify.
Comments