Atlassian CEO confirms no Chinese AI models in production systems
Atlassian chief executive states the company relies on Nvidia and Mistral models, dismissing concerns about Chinese AI while noting customer pushback exists.
Key points
- Atlassian does not use Chinese AI models in its production environment.
- The company utilizes Nvidia Nemotron and Mistral models for its AI tasks.
- CEO Mike Cannon-Brookes argues that rejecting open models is unsustainable.
Atlassian chief executive Mike Cannon-Brookes stated that the company does not run any Chinese AI models in its products. He made this clarification during an interview with Harry Stebbings, host of the 20VC podcast, which was published on Monday following recording in Amsterdam.
The interview took place in the context of Atlassian’s Team ’26 Europe conference held in Amsterdam the previous week. Stebbings raised the topic after SaaStr founder Jason Lemkin reportedly told him that large enterprises are reluctant to adopt Chinese open models. Cannon-Brookes acknowledged that some customer pushback exists regarding Chinese models, but noted there is no similar resistance to open models in general.
Model Selection Strategy
Atlassian relies on Nvidia’s Nemotron models through a deep partnership with the chipmaker, as well as models from Mistral and other providers. Cannon-Brookes explained that the company’s objective is to select the most effective model for each specific task while minimizing costs for customers. He emphasized that Atlassian is very clear about not using Chinese models in production.
Cannon-Brookes argued that the distinction between Chinese and non-Chinese models is often harder to define than it appears. He claimed that many US models are trained partly on Chinese models without disclosing this fact, though he provided no specific examples or evidence to support this assertion. He suggested that refusing open technology that firms can inspect will be difficult to sustain long-term.
Enterprise AI Trends
The CEO expressed skepticism toward companies that claim to have built their own proprietary AI models. He stated that most of these firms have simply taken a Chinese open-weight model and fine-tuned it for a single job. Atlassian also fine-tunes open models, but Cannon-Brookes said they do not label them as their own creations.
He noted that the benefits of building specialized models remain uncertain. For Cannon-Brookes to change his view, specialized models would need to consistently outperform general models on both price and performance. He observed that few current specialized models meet this bar, and many early startups have been overtaken by larger foundation models.
Cost was another significant theme in the discussion. Cannon-Brookes said that large customers are now highly concerned about the expenses associated with AI. Atlassian has a dedicated AI gateway team that routes each task to the most suitable model, but he advises product teams not to focus on cost too early in the development process.
What should you do?
Security and technology professionals should monitor vendor disclosures regarding the origins of AI models used in enterprise software. Organizations should review their own policies on open-weight models and assess whether they require transparency about training data sources. Watch for updates from major vendors like Microsoft and Atlassian regarding their AI partnerships and model selections.
What to do and how to stay safe: Atlassian
- Review vendor documentation to understand which AI models are integrated into your software stack.
- Assess your organization's policy on using open-weight AI models and their training data origins.
- Monitor announcements from major cloud and software vendors for changes in their AI partnerships.
- Evaluate the cost implications of AI usage in your enterprise applications.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Does Atlassian use Chinese AI models in its products?
No, Atlassian CEO Mike Cannon-Brookes stated that the company does not run any Chinese AI models in production.
Which AI models does Atlassian currently use?
Atlassian uses Nvidia’s Nemotron models and models from Mistral, among others, to handle various tasks.
What is Atlassian's stance on specialized AI models?
Cannon-Brookes is skeptical of specialized models, stating they rarely beat general models on price and performance.
