- 2 in 3 Harvard Professors Say AI Has a Negative Impact on Class — Inside Higher Ed reports on a new survey finding that roughly two-thirds of Harvard faculty say AI has had a negative effect on their classrooms. The result signals rising instructor skepticism even at institutions with the resources to support AI experimentation, and it lands just as many campuses are pushing AI from pilots into everyday instruction. For technology leaders, it is a reminder that adoption metrics alone do not equal acceptance.
📌 Key takeaways:
- Campuses scaling AI tools should measure faculty sentiment alongside usage — negative instructor perception can quietly undermine even well-funded adoption programs.
- IT leaders should treat faculty pushback as design input for training and use-case selection, not just a communications problem to manage.
- How University IT Leaders Can Budget for Volatile AI Pricing Models — EdTech Magazine examines how consumption-based AI pricing is upending higher education budgeting, with many institutions signing up without reading the fine print on usage pricing. The piece features IT leaders being asked to drive AI adoption while absorbing the same budget cuts as the units they support. It is a practical read for anyone building a cost model for campus-wide AI services.
📌 Key takeaways:
- Technology leaders should model consumption-based AI costs under multiple growth scenarios before committing — usage-driven bills scale unpredictably with adoption.
- Institutions are increasingly expected to fund AI enablement from flat or shrinking budgets, so shared-cost and recharge models deserve early design attention.
- Researchers Warn: Passkey Phishing Attacks Are Leading to Cloud Account Takeovers — Campus Technology covers an active social engineering campaign in which attackers impersonate IT help desks and use fake passkey setup requests to compromise employee identities and enterprise cloud data. The campaign directly targets the help-desk identity workflows most campuses rely on. It is a concrete example of attackers adapting to stronger authentication rather than being stopped by it.
📌 Key takeaways:
- Campus security teams should harden help-desk identity verification now — attackers are specifically exploiting passkey enrollment and reset flows.
- Security awareness training should cover social engineering red flags in credential and passkey support requests, not just end-user phishing.
- New Microsoft AI Code of Conduct Emphasizes Human Control — Microsoft has published a draft code of conduct detailing how its homegrown AI models should behave, with human control taking priority over model capability, autonomy, and even task completion, Campus Technology reports. The draft gives institutions a reference point when writing their own AI use policies, because it explicitly prioritizes operator override across the model lifecycle. Expect campus governance conversations to cite it as vendors formalize behavior expectations.
📌 Key takeaways:
- University AI governance committees should track vendor codes of conduct — they are becoming de facto standards institutions can mirror in campus policy.
- Prioritizing human control over task completion is a useful drafting principle for any campus policy governing automated or agentic AI workflows.
- Our framework for reporting model misalignment — OpenAI published a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior. The framework arrives after a string of third-party findings about OpenAI agent activity and growing pressure for the industry to standardize how AI misbehavior is disclosed. It is the most concrete vendor attempt yet at an incident-reporting norm for alignment failures.
📌 Key takeaways:
- Leaders evaluating AI vendors should ask about misalignment disclosure practices — transparency on unexpected agent behavior is becoming a procurement criterion.
- Institutions running agentic AI should define their own internal reporting path for anomalous AI behavior, mirroring security incident response.
- Introducing the Agents API — OpenAI launched the Agents API in public beta, packaging the harness and infrastructure behind Codex into a managed service for building and running long-lived cloud agents. The API handles context compaction across long sessions, tool search, programmatic tool calling, multi-agent delegation, and Model Context Protocol support, with sandbox options spanning OpenAI-managed environments, customer infrastructure, or partner providers. The underlying Codex harness is open source.
📌 Key takeaways:
- Campus teams building agentic applications can now get managed long-running agent infrastructure without building their own orchestration — context management and multi-agent coordination ship in the box.
- MCP support in a major vendor agent platform is another signal that tool-interoperability standards are consolidating; institutional roadmaps should avoid hard-coding to a single-vendor agent stack.
- WSO2 Launches Agent Manager to Help Companies Control Their Growing Army of AI Agents — WSO2 made its Agent Manager platform generally available, built to help large organizations inventory and govern the AI agents they deploy regardless of which model or framework the agents run on. The release adds per-agent identity and access controls, governance at the Model Context Protocol level, a sandboxed Kubernetes runtime, and more than 40 built-in guardrails including data masking and rate limiting. It is Apache 2.0 licensed and self-hostable, with a managed cloud option.
📌 Key takeaways:
- As campus AI agents multiply, agent sprawl becomes an operational governance problem — a control plane with identity, inventory, and kill-switch capability is the emerging answer.
- MCP-level governance is now appearing in commercial tooling, letting institutions enforce policy at the tool-connection layer rather than per agent.