Microsoft is rolling out an unorthodox pricing model for its new security chatbot that becomes available to the public on April 1.
As part of a swarm of generative artificial intelligence announcements last year, Microsoft introduced a preview last March of Copilot for Security, which taps large language models to help cybersecurity professionals understand critical issues.
On Wednesday, Microsoft said it will use a consumption-based model, charging $4 per “security compute unit.” Andrew Conway, vice president of security marketing at Microsoft, said the types of prompts and summaries will vary dramatically in size, depending on the customer and type of workload.
“Customers can buy what they need, and that can easily be changed over time without friction,” Conway said in a statement.
Security is a significant business for Microsoft, accounting for more than $20 billion in revenue in 2022, making it larger than gaming or search advertising at the time. Gaming is now bigger with the acquisition late last year of Activision Blizzard.
Microsoft has broadly been working to add generative AI from OpenAI into Windows, Dynamics business applications and other products. Wall Street has been eager to see how Microsoft will be able to make money from AI after investing billions of dollars in OpenAI and AI-related data center equipment.
The pricing for Copilot for Security is designed to keep expenses low for organizations that experiment with the tool while scaling for power users. Microsoft considered input from early customers as well as the costs of tapping OpenAI’s LLMs that process users’ prompts, Vasu Jakkal, a corporate vice president at Microsoft, told CNBC.
Microsoft charges for use of its Azure OpenAI Service based on the number of tokens a client uses. Each token is equal to about four English characters.
It’s a much more convoluted pricing model than other Microsoft tools released of late, such as customer service and general productivity assistants. The Copilot for Microsoft 365 costs $30 per person per month for companies.
BP is an early customer of the new security service. Chip Calhoun, the company’s vice president of cyber defense, said in an email that, “Copilot has made us more efficient and helped us to find attack patterns that could easily be missed without specific use cases.”
Copilot for Security can answer questions by drawing on information from Microsoft’s own security products and third-party providers. It can explain security vulnerabilities, analyze scripts, answer questions about devices and summarize incidents.
Other security software companies dabbling in generative AI include CrowdStrike, which has a chatbot called Charlotte that costs $20 a year per device.
Cyberattacks are becoming a bigger threat by the day. Microsoft said in January that a Russian intelligence group had accessed some of its executives’ email accounts. Roku and UnitedHealth also said they were hit by attacks this year.
Microsoft CEO Satya Nadella said on the company’s most recent earnings call that the latest spate of cyberattacks “highlighted the urgent need for organizations to move even faster to protect themselves from cyberthreats.”