OpenAI has launched a beta version of Developer mode for ChatGPT, which allows full read and write support for Model Context Protocol tools. This new feature, while promising enhanced automation capabilities for developers, comes with significant warnings about its potential dangers. Edwin Arbus, OpenAI’s developer community lead, highlighted that developers can now create connectors to perform actions such as updating Jira tickets or triggering workflows in services like Zapier. User feedback has been largely positive, with many expressing that full Model Context Protocol support could streamline integrations and workflows considerably. However, there are reported bugs, including instances where the system returns inaccurate information. Concerns have also been raised about security risks, particularly regarding prompt injections and the possibility of malicious users exploiting these new capabilities to access sensitive information. As the industry rapidly adopts Model Context Protocol, the urgency for secure configurations cannot be understated, especially given the potential for significant data loss or breaches.
OpenAI has announced the release of GPT-5-Codex, its latest large language model optimized specifically for coding tasks. This new version has reportedly led to a tenfold increase in usage among developers in just one month, thanks in part to its integration with GitHub for dynamic code reviews, which automatically checks for bugs and compatibility issues. The Codex model has been trained on real-world coding tasks, including building projects from scratch and performing large-scale refactoring, making it a powerful tool for software engineers. According to Aaron Wang, a senior software engineer at Duolingo, Codex excelled in identifying complex bugs that other tools missed, thus demonstrating its potential for improving coding efficiency and quality.
Microsoft is introducing free Copilot Chat features to its Office applications for all Microsoft 365 business users. The new update includes a Copilot Chat sidebar in key applications like Word, Excel, PowerPoint, Outlook, and OneNote, allowing users to draft documents and analyze spreadsheets without needing an additional Copilot license. Seth Patton, General Manager of Microsoft 365 Copilot product marketing, emphasized that this secure AI chat is content-aware, tailoring responses to the document currently in use and is included at no extra cost. While users without the paid Microsoft 365 Copilot license will benefit from these features, the premium version, priced at $30 per user per month, offers enhanced capabilities like file uploads and the latest technology advancements, including faster response times and improved availability. This update follows Microsoft’s earlier introduction of AI features into consumer Microsoft 365 plans, which occurred alongside a price increase for subscriptions.
Oh, and if you think we know to pay for all this… think again. According to Jo Liversidge, a Vice President at Gartner, AI software licensing is currently in disarray. Speaking at the firm’s Symposium event in Australia, Liversidge highlighted that many major vendors have yet to incorporate artificial intelligence into their contracts, leaving buyers to navigate complex legal documents to understand their purchases. For instance, she noted that Amazon Web Services only addresses AI in a specific clause of its supplementary terms and conditions. Liversidge also pointed out the inconsistent pricing strategies among AI vendors. For example, Salesforce launched its Agentforce platform with one pricing model and later introduced an alternative, allowing customers to choose between them. Furthermore, some vendors use a credit-based pricing model that can lead to unexpected costs, and they often lack the tools to help customers manage these expenses effectively. This chaotic landscape underscores the need for buyers to exert pressure on vendors to include responsible AI principles in contracts, as only one percent of vendors currently do so.
Why do we care?
AI shifts from optional add-on to default business infrastructure.
OpenAI’s Developer Mode cracks open ChatGPT to act directly inside workflows—Jira, Zapier, whatever you wire it to. That’s powerful, but it’s also a fresh attack surface. You’re inviting trouble if you’re not thinking about secure configurations, prompt injection defenses, and access controls. The real service play here is hardening these integrations, not just turning them on.
GPT-5-Codex is also disruptive. Developers are already considering it because it finds bugs other tools miss. That means customers will start demanding higher coding quality and faster delivery cycles. Providers can’t just resell access—they need to wrap Codex into managed dev, QA, and refactoring services. It’s about proving you can help customers deliver software more reliably.
Microsoft putting Copilot Chat into Word, Excel, and Outlook—for free—changes the baseline. Suddenly, AI is in the hands of every employee, whether IT planned for it or not. That’s Shadow IT at scale. The opportunity isn’t reselling $30 Copilot licenses, it’s helping customers manage usage, secure data flows, and train staff so AI actually delivers productivity rather than risk.
And Gartner’s warning on AI licensing? It validates what most of us already feel: the contracts are inconsistent, the pricing models are unpredictable, and vendors are happy to pass risk downstream. If only one percent of contracts even mention responsible AI, customers need someone to translate, negotiate, and protect them. That’s a consulting and compliance revenue stream, not a resale one.
Stack these together and the message is clear: the value isn’t in licenses. It’s in securing integrations, extracting productivity, managing contracts, and shielding customers from unpredictable costs and risks. AI is now table stakes. Providers win by being the ones who make it usable, safe, and cost-controlled.