OpenAI announces network security plan: Spending millions of dollars, AI may automatically patch network vulnerabilities in the future

Produced by Sohu Technology

Author｜Zheng Songyi

On June 2, Beijing time, the U.S. OpenAI artificial intelligence research company announced the Cybersecurity Grant Program, which will cost US$1 million to enhance and quantify AI-driven cybersecurity capabilities and promote high-level AI Cybersecurity discussion.

Recently, network security problems have occurred frequently due to the abuse of AI technology. The predictions of Musk and others about the "AI security crisis" have been fulfilled ahead of schedule. It is understood that there have been many high-tech fraud incidents across the country recently. Criminals used "AI face-changing" and "AI voice imitation" technologies to pretend to be the victim's friends, and defrauded the amount of up to one million yuan in 9 seconds.

Industry experts have analyzed that AI technology itself also has security vulnerabilities. The current mainstream deep neural network and other technologies have "black box" properties, which will cause some AI algorithms to be uninterpretable. This also means that AI technology has a lot of problems in algorithm security. There are uncertainties in the industry, which may lead to safety hazards and risks in the implementation of industrial applications.

OpenAI’s official website states, “Our goal is to promote the development of artificial intelligence-based cybersecurity capabilities for defenders through grants and other support.”

According to reports, the plan contains three main goals:

1. Empower defenders: We hope to ensure that the most cutting-edge AI features benefit defenders first.

2. Measuring Capabilities: Work to develop methods to quantify the cybersecurity capabilities of AI models in order to better understand and improve their effectiveness.

3. Elevate the discussion: Committed to promoting rigorous discussion at the intersection of AI and cybersecurity, encouraging a comprehensive and nuanced understanding of the challenges and opportunities in the field.

OpenAI stated, “Conventional wisdom in cybersecurity holds that this environment naturally gives attackers an advantage over defenders. An old axiom sums it up: ‘The defense must be 100% correct, the attacker only has to be correct Once.' While it may be true that attackers face fewer constraints and take advantage of their flexibility, defenders have something more valuable - coordination to achieve the common goal of keeping people safe."

It is reported that OpenAI’s current ideas for the plan include but are not limited to allowing AI to automatically patch network vulnerabilities, automatically identify security issues in network source code, and use deception technology to trap attackers.

OpenAI has now opened its project recruitment plan. OpenAI will evaluate and accept applications for funding or other support, with priority given to practical applications of artificial intelligence in defensive cybersecurity (tools, methods, processes). OpenAI will provide incremental funding of $10,000 in the form of API credits, direct grants, and/or equivalents from the $1 million fund.

OpenAI emphasized that it will not currently consider funding offensive security projects.

The above is the detailed content of OpenAI announces network security plan: Spending millions of dollars, AI may automatically patch network vulnerabilities in the future. For more information, please follow other related articles on the PHP Chinese website!