OpenAI Bolsters Ecosystem with “Smoother” GPT-5.3 Model and Strategic Security Acquisition

Prior to the acquisition news, OpenAI launched GPT-5.3 Instant on March 3, 2026, targeting a more “natural, accurate, and direct” user experience.

RMN News Technology Desk
New Delhi | March 10, 2026

In a rapid succession of major moves in early 2026, OpenAI has announced the acquisition of AI security firm Promptfoo and the official release of GPT-5.3 Instant, signaling a dual focus on enterprise-grade safety and more natural user interactions. These developments come alongside significant strategic milestones, including a landmark partnership with Amazon and a major deal with the Pentagon.

Securing the Enterprise Frontier

On March 9, 2026, OpenAI revealed its intent to acquire Promptfoo, a leading platform dedicated to identifying and remediating vulnerabilities in AI systems. The acquisition is specifically designed to accelerate the security and testing capabilities of OpenAI Frontier, a platform used for operating “AI coworkers” within professional workflows.

Once the deal closes, OpenAI plans to integrate Promptfoo’s technology directly into Frontier, providing developers with native tools for automated red-teaming. These tools aim to prevent common AI risks such as:

• Prompt injections and jailbreaks.
• Data leaks and tool misuse.
• Compliance issues related to governance, risk, and compliance (GRC) expectations.

Promptfoo, led by founders Ian Webster and Michael D’Angelo, is currently trusted by over 25 percent of Fortune 500 companies. OpenAI has committed to maintaining Promptfoo’s widely used open-source CLI and library alongside its new integrated enterprise features.

A More Human Conversational Experience: GPT-5.3 Instant

Prior to the acquisition news, OpenAI launched GPT-5.3 Instant on March 3, 2026, targeting a more “natural, accurate, and direct” user experience. This update addresses long-standing user feedback regarding the “preachy” tone and excessive disclaimers found in previous models.

Key features of the new model include:

• Refined Judgment: The model provides direct answers without the defensive “dead ends” or lengthy moralizing preambles that previously interrupted conversations.
• Reduced Hallucinations: Internal evaluations show a 26.8% reduction in hallucinations when using the web and a 19.7% reduction when relying on internal knowledge.
• Texture-Rich Writing: For creative tasks, the model is described as having more range and texture, favoring observed detail over abstract sentiment.
• Smarter Web Integration: Rather than providing cluttered lists of links, the model now synthesizes real-time web data with its internal knowledge to provide contextualized answers.

While the update is available now for all ChatGPT users, OpenAI noted that work continues on improving performance in non-English languages like Japanese and Korean, where the tone can still feel stilted.

Strategic Landscape in 2026

These announcements are part of a broader aggressive expansion by OpenAI in early 2026. This period has been marked by a massive $50 billion partnership with Amazon to revolutionize enterprise AI and a joint statement with Microsoft regarding their ongoing collaboration.

As OpenAI integrates Promptfoo’s security expertise into its development workflows, the company is positioning itself to meet the rigorous demands of high-stakes industries like medicine, law, and finance, where reliability and safety are paramount. For current users, the transition to the new model is underway; while GPT-5.2 Instant remains available as a legacy model for now, it is scheduled for official retirement on June 3, 2026.