Key Developments

OpenAI released GPT-5.4 on March 5, 2026, marking a significant leap in AI capabilities with a context window of up to 1,000,000 tokens—roughly 50-100 times longer than previous models. This “most capable and efficient frontier model for professional work” can handle long-running tasks and multi-application workflows with fewer back-and-forth interactions.

Alongside the model release, OpenAI launched Codex Security, which analyzes software codebases to identify real vulnerabilities while reducing false positives by over 90%. In beta tests, it discovered critical issues like cross-tenant authentication bugs that traditional tools missed.

Meanwhile, significant policy developments emerged across Europe and the US. The European Commission published the second draft of its Code of Practice on AI-generated content marking and labelling, advancing AI Act implementation. The EU also announced a €75 million EURO-3C Project to build federated infrastructure for digital sovereignty.

Industry Context

The massive context window breakthrough addresses a fundamental limitation in AI applications. Previously, developers struggled with maintaining context across complex, multi-step workflows. GPT-5.4’s ability to “steer” output by planning steps mid-response could transform how businesses automate Excel tasks, presentations, and research workflows.

The parallel policy developments reflect growing tensions between innovation and regulation. Anthropic’s lawsuit against the US Defense Department—after being labelled a supply-chain risk for refusing mass surveillance applications—has drawn support from over 30 OpenAI and DeepMind employees, highlighting industry concerns about government overreach.

Practical Implications

For Irish and European developers, these developments create new opportunities while raising compliance questions. The expanded context windows enable more sophisticated applications, from automated financial reporting to comprehensive code analysis. However, the evolving EU marking requirements mean builders must consider how to identify AI-generated content in their applications.

The Snowflake-OpenAI $200 million partnership demonstrates enterprise demand for “sovereign” AI solutions—keeping sensitive data within secure perimeters while accessing advanced capabilities.

Open Questions

Key uncertainties remain around implementation costs for the extended context windows, specific EU marking requirements, and how policy tensions might affect access to cutting-edge models. The rapid revenue growth at Anthropic—reportedly on track to surpass OpenAI by end-2026—also raises questions about market consolidation and competitive dynamics.

Source: OpenAI News