OpenAI's Amp agents now tackle non-code workflows via GPT-6 Astra, Ultrafast and new integrations

OpenAI published first-hand research detailing how its Amp platform extends AI agent capabilities beyond traditional coding tasks. According to the company’s own documentation, this expansion leverages GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT to broaden operational scope.

What new models are integrated into the Amp platform?

The research highlights integration with GPT-6 Astra, Ultrafast, and GPT-Live-1. OpenAI also notes the inclusion of Sign in with ChatGPT as part of this updated agent framework. These models enable Amp agents to execute tasks that do not involve writing or debugging code.

How does this change the scope of agent operations?

Previously focused on development workflows, Amp agents now operate in broader contexts. The published material indicates that these agents can handle non-code tasks by utilizing the newly supported model suite. This shift suggests a move toward general-purpose automation within the platform.

Which specific technologies power these expanded capabilities?

The expansion relies on GPT-6 Astra and Ultrafast for processing, alongside GPT-Live-1 for real-time interactions. Sign in with ChatGPT provides the authentication layer for these agent interactions. OpenAI states that this combination pushes the frontier of what agents can accomplish.

What are the implications for security operations teams?

Security teams should monitor how these agents interact with internal systems. The ability to perform non-code tasks may increase the attack surface for data exfiltration or unauthorized actions. Organizations using Amp should review agent permissions and model access controls immediately.

Questions to ask your suppliers

Teams should verify which specific models their suppliers use for automated tasks and whether those models access sensitive data. Check if the supplier has defined clear boundaries for agent actions outside of code generation. Determine who within the organization receives alerts when agents perform non-standard operations. Ensure that authentication methods like Sign in with ChatGPT are properly scoped to prevent privilege escalation. Review whether the supplier provides visibility into agent decision-making processes for audit purposes.

Background: Attack surface

Attack surface analysis is the process of identifying every digital interface where an attacker might interact with your systems. It goes beyond traditional vulnerability scanning to map external-facing assets, shadow IT, and exposed APIs. This method helps you prioritize risks by understanding how these interfaces connect to critical data.

Read the full guide: Attack Surface Analysis: Map Hidden Entry Points Before Exploitation

What to do and how to stay safe: OpenAI

• Review current agent permissions to ensure they align with the principle of least privilege.
• Monitor logs for any unusual agent activity that falls outside expected non-code workflows.
• Verify that authentication mechanisms for agent interactions are robust and properly configured.
• Assess whether existing security controls can detect and block unauthorized agent actions.

Step-by-step guide: AI Usage Policies at Work: How to Prevent Data Leakage

General security guidance from the TechSec Feed newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which models are now supported by OpenAI's Amp agents?

Amp agents now support GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.

Do Amp agents still focus primarily on coding tasks?

No, the research shows Amp agents are now used for tasks beyond code.

Is this an official release or just research?

This is first-hand research published by OpenAI demonstrating new capabilities.