Recorded Future Inc. today launched a cybersecurity tool that uses an artificial intelligence model from OpenAI LP to detect threats.

Boston-based Recorded Future provides a software platform that helps companies track hacker activities. A bank, for example, can use the platform to detect new malware campaigns targeting the financial sector. Recorded Future says that its technology is used by more than half of the Fortune 100.

Insight Partners acquired a majority stake in the company three years ago. The deal valued it at more than $780 million.  

Recorded Future AI, the new tool the company debuted today, is based on a neural network from OpenAI’s GPT series of large language models. The GPT series includes GPT-4, the advanced neural network that debuted last month. The product line also includes more than a half dozen other AI models with different feature sets.

Companies detect breaches by continuously collecting data about the user activity, applications and hardware in their networks. Historically, cybersecurity teams analyzed that data manually to find malicious activity. Recorded Future AI is designed to ease the task.

According to the company, its new tool automatically identifies breach indicators in a company’s network and prioritizes them based on severity. It also spots vulnerabilities. The tool can, for example, detect if a server contains a configuration flaw that allows users to log in without a password.

Recorded Future AI promises to speed up a number of other tasks as well.

As part of their work, cybersecurity teams regularly produce reports for executives that detail how well the corporate network is protected and what can be improved. To create such a report, analysts must manually gather technical data from a variety of systems. Recorded Future AI promises to automate parts of the process and thereby speed it up by up to several days.

“Now, with Recorded Future AI, we believe we can eliminate the cyber skills shortage and increase the capacity for cyber readiness by immediately surfacing actionable intelligence,” said co-founder and Chief Executive Officer Christopher Ahlberg.

Recorded Future developed the tool by training the GPT model it sourced from OpenAI on 100 terabytes of cybersecurity data. The data was collected by the startup’s namesake software platform. The platform provides companies with intelligence about cyberattacks, vulnerabilities and the servers that hackers use to run malware campaigns, as well as related information. 

The tool also draws on data from the startup’s Insikt Group research team. In particular, it incorporates 40,000 analyst notes about online threats that the Insikt Group has produced over the years. Such analyst notes are used by cybersecurity teams to explain hacker tactics and share related technical information.

Recorded Future is one of several companies using OpenAI’s GPT series of language models to support cybersecurity teams’ work. 

Last month, Microsoft Corp. introduced a service called Security Copilot that incorporates OpenAI’s latest GPT-4 model. The service automatically detects malicious activity and anticipates what steps a hacker is likely to take next during a breach attempt. Cybersecurity teams can use the data provided by Security Copilot to inform their breach remediation efforts. 

Image: Recorded Future