Get the latest on global AI regulations, legal risk, and safety-by-design strategies. Read the Report
Protect your AI applications and agents from attacks, fakes, unauthorized access, and malicious data inputs.
Control your GenAI applications and agents and assure their alignment with their business purpose.
Proactively test GenAI models, agents, and applications before attackers or users do
The only real-time multi-language multimodality technology to ensure your brand safety and alignment with your GenAI applications.
Ensure your app is compliant with changing regulations around the world across industries.
Proactively identify vulnerabilities through red teaming to produce safe, secure, and reliable models.
Detect and prevent malicious prompts, misuse, and data leaks to ensure your conversational AI remains safe, compliant, and trustworthy.
Protect critical AI-powered applications from adversarial attacks, unauthorized access, and model exploitation across environments.
Provide enterprise-wide AI security and governance, enabling teams to innovate safely while meeting internal risk standards.
Safeguard user-facing AI products by blocking harmful content, preserving brand reputation, and maintaining policy compliance.
Secure autonomous agents against malicious instructions, data exfiltration, and regulatory violations across industries.
Ensure hosted AI services are protected from emerging threats, maintaining secure, reliable, and trusted deployments.
Udemy, a global leader in online learning, needed a scalable solution to maintain course quality. Learn how they use ActiveFence to manage content and build user trust.
Udemy's marketplace of 210,000 courses includes the Udemy Business platform, which makes over 27,000 courses in 16 languages available to millions of learners across the world. Since its inception, Udemyโs Director of Curation, Jasmin Tokatlian and her team of three curation managers, had the responsibility of handling all content moderation for the Udemy Business platform. Due to their limited tooling and resources, the team was unable to proactively assess course quality, and spent most of their time responding to client escalations and handling items on a case-by-case basis.
This approach quickly became infeasible to manage and unscalable. With a growing customer base of over 16,000 global organizations, Tokatlian needed to find a way to proactively ensure that the courses offered met her clientsโ expectations for quality and appropriateness.
โWe needed a much more sophisticated and rigorous content moderation philosophy. We could either build this in-house, or find somebody else who knew what they were doing.โ
To support Udemyโs increased scale while ensuring a positive learner experience, Tokatlian looked for a partner who could support the team as it scaled, both in detecting harmful content, but also in training the team on Trust & Safety practices. Tokatlian reached out to ActiveFence for its automated detection capabilities, range of content moderation tools, and vast experience in the Trust & Safety space.
โThe technology was far more sophisticated than we had seen with other companies.
After a short integration and calibration period, ActiveFence implemented its automated detection capabilities to scan course transcripts and material to assess their appropriateness for work environments and ensure they meet professional standards.
โTo be more efficient, we started scanning transcripts of our courses at a massive scale, that human beings just cannot do. ActiveFence really saved the day because not only does it have the technology, but the technology has the ability to learn and grow with us.โ
Aligning to Udemyโs specific business needs was crucial to ensure the projectโs success. This included a calibration process to align and tweak the models to Udemyโs specific metrics to assess content quality, predict learning experience, and minimize profanity and inappropriate content.
โThe core of learning is feedback, growth, and improvement. Those values are inherent in our workplace, so it's really special to find a partner who can also embody those values and work with us in the same way. It's been great.โ
Since implementation, ActiveFence and Udemy continue to tweak and update these models with feedback loops. This has enabled the Udemy team to look into the content at a very granular level, and surface anything that could potentially be problematic before a customer could see it and experience friction.
The partnership with ActiveFence has a profound impact on Udemy's ability to maintain high standards of content quality and appropriateness, leading to improved customer satisfaction and trust.
โOne metric that I track over time is the average rating of all 27,000 courses given by Udemy Business learners. I track that every quarter since I've been at Udemy, and every single quarter that average rating has gone up and up and up and up.โ
By leveraging ActiveFenceโs automated content moderation, Udemy have been able to apply a more proactive, scalable approach to moderation, simultaneously improving user satisfaction and efficiency.
By opting to partner with ActiveFence rather than build their own, the team has been able to focus on its learners and core business needs, instead of responding to content escalations.
Jasmin Tokatlian
Director of Curation
Explore how The Trevor Project leverages ActiveFenceโs tools to create a safe, supportive space for LGBTQ+ youth.
Learn how Cohere partnered with ActiveFence to enhance trust & safety across their platform.
See how Niantic boosts user safety and engagement with ActiveFenceโs cutting-edge technologies.