Close Menu
Versa AI hub
  • AI Ethics
  • AI Legislation
  • Business
  • Cybersecurity
  • Media and Entertainment
  • Content Creation
  • Art Generation
  • Research
  • Tools

Subscribe to Updates

Subscribe to our newsletter and stay updated with the latest news and exclusive offers.

What's Hot

Researchers raise red flags on false images generated by AI in biomedical research

May 9, 2025

Apple develops custom chips such as smart glasses

May 9, 2025

AI Act: Statewide Spotlight – Regulatory Surveillance Podcast | Troutman Pepperlock

May 9, 2025
Facebook X (Twitter) Instagram
Versa AI hubVersa AI hub
Friday, May 9
Facebook X (Twitter) Instagram
Login
  • AI Ethics
  • AI Legislation
  • Business
  • Cybersecurity
  • Media and Entertainment
  • Content Creation
  • Art Generation
  • Research
  • Tools
Versa AI hub
Home»Cybersecurity»Invisible, autonomous and hackable: AI agent dilemma never saw no one coming
Cybersecurity

Invisible, autonomous and hackable: AI agent dilemma never saw no one coming

By February 20, 2025No Comments5 Mins Read
Share Facebook Twitter Pinterest LinkedIn Tumblr Reddit Telegram Email
Share
Facebook Twitter LinkedIn Pinterest Email

This article is part of the special issue of VentureBeat, “Cyber ​​Resilience Playbook: Naviging of Threats.” For more information about this special issue, please see here.

Generated AI raises interesting security questions and as businesses move into the world of agents, their safety issues increase.

When AI agents enter the workflow, they need to have access to sensitive data and documents to do their job, a major risk for many security-oriented companies.

“The increased use of multi-agent systems introduces new attack vectors and vulnerabilities that can be exploited if they are not properly protected from the start.” “However, the impact and harm of these vulnerabilities is , it could be even bigger because the amount of attachment points and interfaces that multi-agent systems have is increasing.”

Why AI agents pose such high security risks

AI agents (autonomous AI that performs actions on behalf of users) have been extremely popular in the past few months. Ideally, you can connect to boring workflows and perform any task, from finding information based on internal documents to creating recommendations that human employees should take .

However, they present an interesting issue to enterprise security experts. They need to access data that makes personal information effective without accidentally opening or sending it to others. As agents do more tasks, human employees once did this, creating issues of accuracy and accountability, and headaches for security and compliance teams.

AWS CISO Chris Betz told BentureBeat that the searched high-generation (RAG) and agent use cases are “a compelling and interesting angle of security.”

“Organisations need to think about what an organization’s default sharing looks like, because agents find things in searches that support their mission,” says Betz. “If you want to overshare documents, you need to think about default sharing policies within your organization.”

Security experts should ask if an agent should be considered a digital employee or software. How much access does the agent need? How should they be identified?

AI Agent Vulnerability

Gen AI makes many businesses more aware of potential vulnerabilities, but agents were able to open them up to more issues.

“Attacks that affect single agent systems, such as data addiction, rapid injection, and social engineering, can affect agent behavior.

Companies need to pay attention to what agents have access to to ensure that data security remains.

Betts noted that many security issues surrounding access to human employees can be extended to agents. So it “results in making people accessible only to the right and the right thing.” He added that “each of these stages is an opportunity” for hackers when it comes to agent workflows with multiple steps.

Give your agent an identity

One answer is to issue a specific access identity to the agent.

He said the world where models infer about problems over several days is “a world where we need to think more about recording the identity of an agent and the identity of the person in charge of that agent anywhere within the organization.” Jason Clinton, CISO of humanity.

Identifying human employees is something businesses have been doing for a very long time. They have a specific job. They have the email address they use to sign the signature and be tracked by their IT administrator. I have a physical laptop with an account that I can lock. They get individual permission to access some data.

This type of employee access and identification variation can be deployed to agents.

Both Betz and Clinton believe this process can inspire enterprise leaders to rethink how they provide information access to their users. It can even lead an organization to overhaul the workflow.

“Using agent workflows actually provides the opportunity to bind use cases for each step along with the data required as part of the RAG, but only the data needed,” Betz said. It’s there.

He added that the agent workflow “helps address some of these concerns about oversharing” as companies need to consider data that has access to full action. Clinton added that in a workflow designed around a particular set of operations, “there is no reason to access the same data that Step 1 requires.”

Old-fashioned audits aren’t enough

Companies can also search for agent platforms that allow them to peek at how agents work. For example, Don Schuerman, CTO at Workflow Automation Provider Pega, said his company can help ensure agents secure by telling users what they are doing.

“Our platforms are already used to audit work that humans do, so we can also audit every step that an agent is doing,” Schuerman told VentureBeat.

Pega’s latest product, AgentX, allows you to switch to a screen where human users provide an overview of the steps an agent takes. Users can see where they are along the agent’s workflow timeline and get a read out of a particular action.

Audit, timeline, and identification are not the perfect solution to the security issues presented by AI agents. However, as companies begin to explore the potential of agents and deploy them, more targeted answers may emerge as AI experiments continue.

author avatar
See Full Bio
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
Previous ArticleGeForce RTX 5070 Ti Powers AI and content creation
Next Article Rethinking content creation Azure AI-powered track story bot

Related Posts

Cybersecurity

Agent AI, IoT Security, Humanoid Robots among Forrester’s Top Emerging Technology in 2025

May 8, 2025
Cybersecurity

AWS surveys show that AI will beat security in the 2025 IT budget

May 8, 2025
Cybersecurity

Commission clears AI security technology trial following knife case in county building

May 8, 2025
Add A Comment
Leave A Reply Cancel Reply

Top Posts

New Star: Discover why 보니 is the future of AI art

February 26, 20255 Views

AI image generation using Flux models: WebUI Forge Quick Start

November 23, 20243 Views

Will AI apps help carry the mental load of moms?

May 8, 20252 Views
Stay In Touch
  • YouTube
  • TikTok
  • Twitter
  • Instagram
  • Threads
Latest Reviews

Subscribe to Updates

Subscribe to our newsletter and stay updated with the latest news and exclusive offers.

Most Popular

New Star: Discover why 보니 is the future of AI art

February 26, 20255 Views

AI image generation using Flux models: WebUI Forge Quick Start

November 23, 20243 Views

Will AI apps help carry the mental load of moms?

May 8, 20252 Views
Don't Miss

Researchers raise red flags on false images generated by AI in biomedical research

May 9, 2025

Apple develops custom chips such as smart glasses

May 9, 2025

AI Act: Statewide Spotlight – Regulatory Surveillance Podcast | Troutman Pepperlock

May 9, 2025
Service Area
X (Twitter) Instagram YouTube TikTok Threads RSS
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer
© 2025 Versa AI Hub. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.

Sign In or Register

Welcome Back!

Login to your account below.

Lost password?