Cal.com moves its core product to closed source, citing AI threats to open source

Cal.com, one of the more prominent open-source players in scheduling, is taking its commercial product closed source. The reason is not a change of business model as such, but a new balance of risk: the company believes modern AI tools have made it far too easy to find vulnerabilities in public repositories, and have made open code too costly to defend when it comes to protecting user data.

The company announced the decision on 15 April 2026. Production code and all further development of the commercial product become private, while the public repository turns into a separate community branch called Cal.diy. That version remains self-hosted and open source, but now under the MIT license instead of the previous AGPL. It keeps the scheduling engine, booking logic, app store and API v2, which is everything individual developers and small teams need to deploy it themselves.

Cal.diy drops what belongs to the managed commercial service and enterprise scenarios: organizations and teams, routing forms, automated workflows, instant booking, AI phone, SAML/SSO, analytics dashboards, some admin interfaces and the old API v1. In effect, Cal.com is splitting the product onto two tracks: an open platform for hobbyists and self-hosting, and a closed production version in which the company takes responsibility for sensitive customer data. Internally, the team puts it simply: they want to be a scheduling company, not a cybersecurity company.

Cal.com's argument rests on the claim that AI radically changes the economics of offensive security. Finding exploitable bugs used to require an experienced researcher and a lot of manual time; now models and agentic tools can work through a codebase step by step, trace data flows, look for business-logic violations and quickly assemble a proof of concept.

In a post on its own blog, Cal.com cites tests by its partner Hex Security: across 28 companies, their autonomous pentest agents found about 2000 vulnerabilities, of which 44.6% were critical or high severity, and 65.1% of scans turned up at least one critical bug. The same post cites a benchmark in which access to source code improved vulnerability detection by roughly 20% compared with black-box testing. As a public symbol of this new reality, the company points to an AI model called Mythos, which, by its account, was able to identify an old vulnerability in BSD and quickly build a working exploit.

Cal.com itself does not claim that closed code automatically makes a product secure. In its official explanation, the company states plainly that this is not a perfect solution and that security through obscurity does not save you on its own. But for a service that handles data about meetings, calendars and bookings, even a small reduction in the likelihood of mass auditing by attackers looks justified. The team also promises to keep security patches flowing back between the closed product and Cal.diy, and says it would like to return to an open model one day if the threat landscape becomes more manageable.

This is a particularly notable reversal, because Cal.com built its brand precisely as an open-source alternative to Calendly and did not hide that it saw open code as a source of growth. According to its leadership, under a different configuration of risks the company would have preferred to stay with its previous model. So this is not a complete break with the community, but an attempt to separate the experimental and commercial parts of the product into different security perimeters.

The main takeaway is broader than the fate of one company. The Cal.com story shows that AI is changing not only software development but the social contract of open source itself: transparency still helps defenders, but it now scales the work of attackers just as well. For commercial open-source products, this may mean a painful choice between the ideology of openness and the duty to reduce risk for customers. Cal.com is the first to make that choice so demonstratively, and that is why its case matters for the whole industry.