Perplexity adds new safeguards and controls for Comet Assistant amid security concerns

Written by: Mane Sachin

Published on:

Follow Us

Comet Assistant Upgrades for Safer, More Controlled AI Browsing

Perplexity has introduced a series of upgrades to its Comet Assistant, focusing on new safeguards and giving users more control over how the AI browser agent handles “high-stakes decisions.”

The updated Comet Assistant will now pause and ask for user approval before performing important actions, such as logging into a website or completing a purchase, the AI company announced in a blog post on November 14.

According to Perplexity, the Assistant operates like a human assistant, handling minor tasks independently while seeking permission for more critical actions before proceeding on behalf of the user.

Users will also gain better visibility into the Assistant’s activities, with the ability to track its reasoning step-by-step in the Assistant sidecar. New interface elements, like clear stop buttons and guidance options, give users more control over its behavior.

The company emphasized that these updates align with three core principles: transparency, user control, and sound judgment. Each improvement is designed to show exactly what the Assistant is doing, ask for input on important decisions, and demonstrate careful judgment before executing sensitive actions.

Enhanced Features, Security, and Future Developments

Perplexity has also upgraded its search omnibar, allowing users to decide whether to browse manually, let Comet Assistant browse once on their behalf, or allow the Assistant to automatically act when it identifies useful opportunities.

These enhancements arrive amid a growing race among AI startups to develop AI-powered browsers to expand user reach and distribution, though such tools have raised concerns around privacy and security.

Security experts have warned that agentic AI browsers, including Comet and OpenAI’s ChatGPT Atlas, could be manipulated through indirect prompt injection, potentially exposing sensitive data like emails, banking information, and personal accounts.

OpenAI has similarly noted that such agents can be vulnerable to hidden malicious instructions embedded in webpages or emails, which could override intended behavior and lead to data theft or unintended actions.

Perplexity maintains that its updates prioritize user control. By increasing transparency, adapting to user preferences, and requesting permission for sensitive tasks, Comet Assistant balances autonomous functionality with keeping users firmly in charge.

Additionally, Comet Assistant can now switch between tabs for tasks like spreadsheet entry, while improvements to its underlying architecture allow it to better navigate and interact with complex websites. The company also teased the launch of a new AI assistant in the near future.

Also Read:

OpenAI Lets Users Tweak ChatGPT to Stop Overusing Em Dashes: Follow these 5 simple steps

Texas Secures $40B Google Investment to Power the State’s AI Boom

Mane Sachin

My name is Sachin Mane, and I’m the founder and writer of AI Hub Blog. I’m passionate about exploring the latest AI news, trends, and innovations in Artificial Intelligence, Machine Learning, Robotics, and digital technology. Through AI Hub Blog, I aim to provide readers with valuable insights on the most recent AI tools, advancements, and developments.

For Feedback - aihubblog@gmail.com