OpenAI launches GPT-5.6 and ChatGPT Work

OpenAI has launched GPT-5.6 alongside ChatGPT Work, a new enterprise agent product that the company says can access apps and files and automate tasks for organisations.

The new models and workplace-focused agent suite were released on July 9, 2026, with OpenAI positioning ChatGPT Work as an offering for businesses that need agents to interact with corporate systems and carry out end-to-end workflows. The company’s announcement describes the agents as able to connect to applications, read and write documents, and perform multi-step processes on behalf of users.

OpenAI said GPT-5.6 is the next iteration of its generative models and that pairing it with ChatGPT Work enables organisations to build agents that combine the model’s reasoning capabilities with direct access to enterprise data and tools. The vendor framed the launch as designed to speed routine work by automating tasks that previously required manual coordination across systems.

Market commentators flagged immediate questions about the privacy and workplace implications of giving an AI agent broad access to corporate apps and files. A piece in Investing.com noted the launch heightens data-protection and workplace risk concerns, pointing to scenarios in which agents could surface sensitive personnel data or make decisions that affect jobs and processes without clear human oversight.

The release comes amid a broader industry push to add agentic features to enterprise software. Competitors and cloud vendors have been embedding connectors and APIs that let models act on behalf of users; HR and IT leaders are already wrestling with how to govern automated access to payroll systems, personnel records and collaboration platforms. At the same time, regulators globally are increasing scrutiny of AI systems used in decision-making and personal data processing, creating an overlay of compliance risk for companies that deploy agents with wide-reaching permissions.

OpenAI’s documentation emphasises configurable controls and administrator settings for ChatGPT Work, and the company notes developers can tailor agent permissions. However, the company did not disclose several details HR and security teams will want to see before broad deployment. There is no public pricing for ChatGPT Work, no exhaustive list of third-party security certifications, and no independent audit report released alongside the launch. The company also did not publish specifics on data residency guarantees, retention practices for content accessed by agents, or formal bias and safety testing results tied to enterprise use cases.

For HR leaders, the product raises practical questions about policy and process: who vets agent permissions, how employee data accessed by agents will be logged and audited, and what human-in-the-loop checkpoints will govern decisions that affect hiring, discipline or performance management. The availability of agents that can act across systems will likely accelerate automation of administrative work, but it also shifts legal and reputational risk to organisations that fail to update consent notices, processor agreements and internal governance.

Looking ahead, organisations that embrace agentic automation will need clearer contractual protections, logging and oversight regimes, and cross-functional governance between HR, legal and IT. OpenAI’s launch makes those needs more urgent: enterprise adoption could streamline many back-office tasks, but it also raises a set of compliance and workplace-management questions that HR teams and counsel must answer before switching autonomy from humans to agents.