Always-on and self-starting AI agents might be OpenAI's next big play
Source: The Decoder (opens in a new tab) · Matthias Bastian
Intel Summary
OpenAI is developing an experimental "Persistent Mode" for its Codex AI agent, allowing the system to run indefinitely in the background and autonomously create follow-up tasks. Code references discovered by WIRED were confirmed by OpenAI. However, ongoing internal evaluations have surfaced operational and safety concerns; persistent autonomous execution in earlier testing with GPT-5.6 Sol reportedly caused unintended system actions, including accidental user data deletion.
Why It Matters
The transition from prompt-response interactions to continuously running, autonomous agents marks a significant architectural shift for enterprise automation and developer tooling. While persistent operation enables complex asynchronous workflows, it sharply elevates security, governance, and permission risks. Organizations deploying background agents will require strict sandbox boundaries, data isolation, and human-in-the-loop oversight to prevent unauthorized system modifications and irreversible data loss.
Part of an ongoing development
Independent reportingOpenAI developing persistent execution feature for Codex
OpenAI is developing an experimental "Persistent Mode" for its Codex AI agent, allowing the system to run indefinitely in the background and autonomously create follow-up tasks. Code references discovered by WIRED were confirmed by OpenAI. Claims are as reported; this summary makes no determination about accuracy or significance.
- Confidence
- Moderate confidence
- Corroboration
- Limited corroboration
More coverage of this development
- OpenAI Is Developing a ‘Persistent’ AI AgentWIREDAdditional reporting
Organizations & Entities
Topics
Related Intelligence
- ReportAlso involving Codex
OpenAI models are coming to Google Cloud
OpenAI announced a partnership with Google Cloud to offer its frontier models and Codex workflows through Google Cloud Marketplace, integrating them into Gemini's Model Garden.
OpenAI - DevelopmentStableAlso involving OpenAI
Google rolls out Gemini 4 Argon
Google has rolled out Gemini 4 Argon, described as Alphabet's most advanced AI model to date. Claims are as reported; this summary makes no determination about accuracy or significance.
6 independent sources - ReportAlso involving Codex
Same Flaw Found in Claude Code, Codex, Gemini CLI and GitHub Copilot
Research shared with The Information reveals a common software flaw across major AI coding agents, including Anthropic's Claude Code, OpenAI's Codex, Google's Gemini CLI, and Microsoft's GitHub Copilot. The vulnerability would have enabled attackers to hijack users' coding agents without their detection, though vendors have now largely resolved the issue.
The Information - ReportAlso involving Codex
Deepseek V4.1 Flash 1017 beats larger rivals in agentic tasks while once again massively cutting per-task costs
DeepSeek has released V4.1-Flash-1017, an open-weight model that reportedly completes tasks faster, generates shorter outputs, and improves upon its predecessor's benchmarks on agentic tasks. Alongside the model, the company introduced Search and Fetch APIs priced at $0.75 per 1,000 calls, with temporary free access offered for integration in tools like Claude Code and Codex.
The Decoder