Abstract:
According to people familiar with the matter, OpenAI, which is vigorously developing its network security business, has started preliminary negotiations and plans to acquire
Trail of Bits
——A security company that specializes in discovering and fixing vulnerabilities in open source software. The size of the potential deal is currently unknown. OpenAI sometimes pays in stock for its acquisitions of smaller startups, although some acquisitions are for less. People familiar with the matter said that negotiations between the two parties are still advancing and there is a possibility that the transaction will fall through.
The team of this start-up is expected to help OpenAI seize one of the most promising business tracks. This year, OpenAI and its main rival, Anthropic, have gone head-to-head with each other in a race to develop purpose-built AI specifically designed to mine software vulnerabilities.
Trail of Bits, founded in 2012, provides security research, security tools, and helps maintain various open source projects. Information on the official website shows that the company has more than 40 employees, the vast majority of whom are safety engineers. The company did not take the conventional risk financing route: it received about $10,000 in funding from the U.S. Department of Defense in 2015; in 2020, it received a $1 million loan from the U.S. federal Paycheck Protection Program.
In June this year, the start-up has already cooperated with OpenAI to jointly promote a software vulnerability repair project called "Patch the Planet (Patch Global Vulnerabilities)". Trail of Bits engineers use the OpenAI Codex series of tools (including network security-specific models) to find, review, and fix vulnerabilities in popular open source projects such as Python.
Trail of Bits CEO Dan Guido said in a promotional video for the project: "For more than ten years, we have accumulated a large number of partnerships in the open source community and earned the trust of the community. Codex allows our team to deliver work in one day that originally took weeks to complete."
As large model capabilities continue to iterate, network security has become a key direction of major AI laboratories. Independent security researchers say hackers have leveraged publicly available AI tools to launch multiple cyberattacks around the world. Last week, the CrowdStrike report mentioned that attackers used open source AI agents and large models from China to carry out network intrusions.
OpenAI and Anthropic cooperate with relevant arrangements of the White House to allow large enterprises to use their strongest models to find and repair security vulnerabilities in their own software. However, in the version released to the general public, the two companies have castrated these capabilities to prevent malicious elements from using the model to launch attacks.
Network security software vendors such as Palo Alto Networks are also using large models from OpenAI and Anthropic to optimize security products sold to enterprises.
Another sign of OpenAI’s ambitions in the field of cybersecurity: The company recently hired Dali Rajic, an executive who previously worked at Google’s cloud security unit Wiz, as chief revenue officer.
Comments