Perplexity adds new controls, built-in safeguards for Comet Assant amid security concerns | Technology News

Perplexity has announced a set of upgrades to Comet Assant, including new safeguards and enhanced user control for when the AI browser agent is making “high-stakes decisions”.Comet Assant will pause and ask users for permission before proceeding with important tasks such as logging in to a certain site or completing a purchase from your shopping cart, the AI search startup said in a blog post on Friday, November 14.
“The Assant is designed to operate under the same principles a human assant would, making some minor decisions without interrupting you while knowing to ask permission for more important decisions before acting on your behalf,” it added.
Story continues below this ad

Additionally, Perplexity said that users will now be able to closely track the AI browser agent’s actions and see its reasoning step–step in the Assant sidecar. It has also added clear buttons to stop Comet Assant or provide more guidance. The San Francisco-based AI startup said that these upgrades to Comet Assant have been made in line with three core principles, namely: Transparency, user control, and sound judgment.
“Every improvement to the Comet Assant advances these principles. It now shows you exactly what it’s doing, asks how you want it to act, and demonstrates sound judgment before completing sensitive actions,” Perplexity said.
The Perplexity Search or omnibar has also been upgraded so that users can more easily choose whether they want to browse the web, allow Comet Assant to browse on their behalf once, or let the Assant automatically browse for users when it detects the chance to be useful.
Comet Assant’s new features come at a time when AI startups are racing to build AI-powered browsers in an attempt to boost user reach and dribution. However, the rise of AI browser agents has led to data privacy and security concerns.Story continues below this ad

Researchers have pointed out that agentic AI browsers such as Perplexity’s Comet and OpenAI’s ChatGPT Atlas could allow attackers to maliciously instruct the browser agent via indirect prompt injection and gain access to sensitive user data, including emails, banking passwords, and other personal information.
OpenAI itself has acknowledged the possibility of such an attack. “Besides simply making makes when acting on your behalf, agents are susceptible to hidden malicious instructions, which may be hidden in places such as a webpage or email with the intention that the instructions override ChatGPT agent’s intended behaviour. This could lead to stealing data from sites you’re logged into or taking actions you didn’t intend,” the ChatGPT-maker has previously said.
However, firms like Perplexity have continued to ins that users will always be in control of AI browser agents. “ prioritising visibility, adapting to your preferences, and requesting permission before sensitive actions, the Comet Assant works at the speed of autonomy while keeping you firmly in control,” the company said.
Recently, Perplexity said it has upgraded Comet Assant with the ability to switch back and forth between tabs in order to fill out a spreadsheet or for other data entry-related tasks. It also made adjustments to the AI browser agent’s underlying architecture, enabling it to better perceive and interact with complex web environments. The $20 billion-dollar Google challenger also said it will be launching a new AI assant in the coming days.