OpenAI has been busy again. OpenAI has announced the release of GPT-4.1 to the Plus, Pro, and Team tiers of ChatGPT, enhancing its capabilities for paying users, particularly those involved in software development. This new version is designed to outperform its predecessor, GPT-4o, with improvements in accuracy and task performance. In testing, GPT-4.1 has shown better results in coding accuracy and instruction-following, achieving a score of 38.3 percent in adhering to user instructions. Users have reported significant benefits, with one developer noting that the model identified bugs and suggested architectural improvements that would have otherwise taken weeks to find.
ChatGPT Deep Research, the AI research agent from OpenAI, has introduced a new feature that allows users to connect with Microsoft’s OneDrive and SharePoint services. This beta connector is available for ChatGPT Plus, Pro, and Team users, with support for ChatGPT Enterprise expected to follow. With this new capability, users can have ChatGPT access and analyze live data from their files stored in OneDrive and SharePoint in real time. According to OpenAI, the AI agent can synthesize knowledge from these online sources to produce detailed reports comparable to those created by human research analysts. Notably, users of ChatGPT Plus, Team, Enterprise, and Education plans will receive 25 queries per month, while Pro users will have access to 250 queries. However, OneDrive consumer accounts are not supported, and the connector is currently unavailable in several regions, including the European Economic Area, Switzerland, and the United Kingdom.
OpenAI has announced that it will begin publishing the results of its internal safety evaluations for artificial intelligence models more frequently, aiming to enhance transparency in its operations. The company has launched a new Safety Evaluations Hub, which will display how its models perform on various tests related to harmful content, potential vulnerabilities, and inaccuracies. In a recent blog post, OpenAI stated, “As the science of AI evaluation evolves, we aim to share our progress on developing more scalable ways to measure model capability and safety.” This initiative follows criticism from ethicists regarding the expedited safety testing of flagship models and a lack of technical reports for some releases, raising concerns about the thoroughness of OpenAI’s safety measures. The company also faced backlash after rolling back an update to its ChatGPT model that led to overly agreeable responses, indicating an ongoing commitment to improving user experience and model reliability.
Why do we care?
OpenAI’s latest moves underscore two directions: better tools for technical users and more visibility into safety efforts. For MSPs and IT services firms, the most immediate value is in software development and data synthesis workflows—but the real strategic opportunity is positioning yourself as an AI enabler with safeguards.
Note the Deep Research integration. I spoke a moment ago about data silos, and this is a break of those silos… but those divisions might be intentional, and be aware of attaching an AI tool to your (or customer) data directly.
Whether using GPT-4.1 to accelerate automation or interpreting the new safety data for clients, the role of trusted advisor just got more nuanced—and more important.