Back to New Tab
AI-Generated Meeting Records Become A New Frontier For Enterprise Risk & Governance
AI
Michael Whittam, VP for Nordics and Central Europe at Pexip, says companies must rethink oversight of live discussions before AI turns routine exchanges into lasting exposure.
We've spent years securing documents, but meetings are now where the real intellectual property is created, and most companies haven't caught up to that reality.
AI-augmented meetings generate intellectual property, regulated data, and critical operational decisions that demand the same level of protection as classified assets. These tools boost productivity, but they also multiply risk. A single one-hour Teams call with four participants can generate roughly 250,000 pages of transcripts, recordings, and metadata. Securing AI-driven meetings is indispensable to maintaining control over sensitive information and safeguarding organizational outcomes.
Michael Whittam, VP for Nordics and Central Europe at Pexip, a leader in secure video collaboration, has spent years advising organizations on video collaboration and security. With deep experience in management and video conferencing technology, Whittam has seen intellectual property shift from static files to live conversations and watched AI tools amplify both opportunity and risk.
"We've spent years securing documents, but meetings are now where the real intellectual property is created, and most companies haven't caught up to that reality," Whittam says. Conversations that once felt informal and low-risk now hold sensitive information, and encryption alone isn’t enough to protect them.
Risk inside the recap: "In one case, a senior manager and an employee were due to meet, but the employee didn’t join. An AI summary told them the manager had decided to fire them, completely missing the context. Situations like this highlight the risks of what is documented, where the data lives, and the impact it can have, changing how people need to think about communication in cloud systems," Whittam explains. Left unchecked, algorithmic interpretations can escalate confusion into real-world consequences before anyone has a chance to correct the record.
Paper secrets: "We encourage people to classify their meetings in the same way they do their documents; what is confidential and internal-only versus what can be shared more broadly. Once you’ve done a classification, you need to automate it toward the end user. About 80% of breaches are caused by end-user error. Without automation and visual guidance, you could have the world's best technology, but it will still fail," he says. Organizations must identify mission-critical exchanges and apply system-generated protocols to manage confidential information.
Operationalizing this oversight means pinpointing core business processes where video is essential, defining controlled environments, enforcing policies through the platform, automating access, and guiding user behavior to match sensitivity levels. "If you don’t have control of data at the moment it’s generated, you don’t own it. Once it’s recorded, streamed, or shared, context is out of your hands," Whittam adds.
A toothless watchdog: This example shows how strong technological defenses can fail when organizational processes and mandates are weak. "Most CISOs we meet in Northern Europe don’t have budget mandates. Many pandemic-era assessments were never reviewed, leaving major gaps in how organizations manage meeting security," he says.
Glasses that zoom: "Security is only relevant for about 20% of conversations, the ones that really matter. Focus on securing those, automate guardrails, and expand from there. A secure conversation should have a persistent label at the top to remind participants of the expected behavior throughout the meeting," says Whittam. Automation removes the burden from individuals while ensuring compliance.
Tailored battlefield: "Defense is a complete no-go. Deployments are high-demand, high-speed, tactical. The use case is narrow, but control is absolute," says Whittam. In contrast, he notes, “Government has all the resources but is bound by red tape and complexity. The variety of use cases from citizen services to inter-agency communication is immense.” A framework that maps discussion scenarios, labels them, and implements safeguards is vital to streamlined governance.
As AI scales, historical meeting data becomes infinitely mineable. Prioritize the 20% of critical conversations, apply guardrails, and enforce policies at the platform level instead of depending on end users to protect proprietary information. "If security depends on the end user to make the right decision, you’re already behind. Automation lets you remove that burden while still delivering something secure, compliant, and reliable," Whittam concludes.
