HomeWinBuzzer NewsZoom Unveils AI Avatars for Video Chats, Raising Deepfake Concerns

Zoom Unveils AI Avatars for Video Chats, Raising Deepfake Concerns

Zoom has announced AI avatars, offering photorealistic replicas for users. Despite safety measures, the feature is fueling deepfake concerns.

-

Zoom just announced plans to bring custom AI avatars to its platform, offering users the ability to create photorealistic digital replicas of themselves by mid-2025. Unveiled during the Zoomtopia event, the avatars will replicate users’ facial movements and voice based on a brief video, syncing the generated audio with the avatar’s lip movements. The update is part of Zoom’s strategy to expand its AI features, already integrated across various tools in Zoom Workplace.

With custom avatars, the company aims to make asynchronous communication more efficient, enabling people to send video messages without actually being on camera. Chief Product Officer Smita Hashim describes the avatars as a time-saving alternative that could streamline communication across teams. However, the introduction of realistic digital avatars inevitably brings up concerns about deepfake misuse and how easily such technology could be exploited.

Custom Avatars and the Challenge of Deepfake Prevention

Zoom’s AI avatars will mimic users’ upper body and facial expressions, allowing for more lifelike video interactions. While this promises to improve user engagement, the potential for abuse is clear. The technology closely resembles deepfake methods that have been used to generate misleading content online, from fake celebrity videos to scam attempts.

The company says it will include safety measures, like watermarking avatar videos and authentication processes, though it hasn’t disclosed exactly how these precautions will work. Critics point out that watermarks can be bypassed or cropped out, raising doubts about the sufficiency of these measures. Zoom’s responses haven’t quelled worries about how effectively it can keep the tool from being used maliciously.

A rollout of these custom avatars comes as deepfake content has already become widespread. Viral videos featuring fabricated appearances of public figures like President Joe Biden or Kamala Harris have fueled misinformation, while impersonation scams involving deepfake audio and video have cost victims over a billion dollars, according to the FTC.

Legal frameworks are still catching up; although some states have laws addressing AI-aided impersonation, federal legislation is lagging. The White House in September secured a commitment from major AI firms about deepfakes. Adobe, Anthropic, Cohere, Microsoft, OpenAI, and Common Crawl—have agreed with the U.S. government to take steps against the utilization of AI technologies in creating harmful deepfake content, particularly non-consensual and exploitative imagery.

Major tech organizations will responsibly manage their data sourcing processes and implement measures aimed at hindering AI-facilitated sexual abuse. These actions include model evaluations and rigorous testing to avert the generation of injurious content.

Zoom AI Companion Adds Productivity Tools

Beyond custom avatars, Zoom AI Companion 2.0 update introduces a range of new productivity features. The assistant will be accessible in a sidebar across Zoom Workplace, including services like Meetings, Chat, Docs, Calendar, and Mail. It’s designed to simplify daily tasks by offering tools for summarizing conversations, generating tasks from meeting notes, and even providing real-time answers during meetings by pulling in data from different sources, including web searches.

To support these expanded capabilities, Zoom relies on a “federated” AI approach, tapping models from partners like OpenAI, Anthropic, and Meta. The assistant can adapt based on the type of task, whether it’s summarizing a chat or drafting a project proposal. Users will soon be able to extend these AI tools to integrate with third-party services like Jira, ServiceNow, and Google Workspace for more seamless task automation.

Upcoming features also include Zoom Tasks, a project management tool launching in December 2024. The assistant will automatically convert meeting summaries into tasks, helping teams stay organized and follow through on action items.

Customizing AI Companion for Specific Business Needs

The newly announced custom add-on for AI Companion aims to give companies more control over how the assistant interacts with data and integrates with internal systems. Scheduled for release at $12 per user per month, the add-on will allow organizations to fine-tune the AI’s behavior, add specialized terminology, and create custom response templates.

Zoom AI Studio will offer tools to adjust the AI’s settings, enabling companies to expand the assistant’s context by linking it to business-specific data sources. For example, adding a custom glossary can help the AI better understand industry terms, while integrating third-party apps can allow it to perform tasks based on meeting content, such as filling out service requests or updating records.

The customizable approach intends to make the AI more relevant to different industries, where workflows and requirements can vary significantly. However, this raises questions about data security and how well the AI can be controlled to prevent errors or inappropriate use of sensitive information.

Last Updated on November 7, 2024 2:37 pm CET

SourceZoom
Luke Jones
Luke Jones
Luke has been writing about Microsoft and the wider tech industry for over 10 years. With a degree in creative and professional writing, Luke looks for the interesting spin when covering AI, Windows, Xbox, and more.

Recent News

0 0 votes
Article Rating
Subscribe
Notify of
guest
0 Comments
Newest
Oldest Most Voted
Inline Feedbacks
View all comments
0
We would love to hear your opinion! Please comment below.x
()
x