A survey released this week by research firm Verasight found that 89 percent of American adults support requiring companies to publicly disclose the results of safety tests on AI models, and 81 percent back federal authority to block the release of AI systems deemed dangerous. The poll, fielded June 18–19 among 1,690 respondents, delivers a strong bipartisan signal that could shape how Microsoft and other tech giants introduce AI features into widely used products like Windows and Copilot.

What the Poll Actually Found

The Verasight survey, with a margin of error of ±2.4 percentage points, asked two straightforward questions:

  1. Should companies be required to publicly disclose safety-test findings for AI models before release?
  2. Should the federal government have the power to block the release of AI models that pose significant risks?

The answers were unambiguous. Support for public disclosure of safety tests reached 89 percent, while 81 percent backed blocking risky releases. Both positions commanded majority support across every demographic group—party affiliation, age, region, and education level—though the intensity varied. Republicans, independents, and Democrats alike signaled they want more transparency and stronger safeguards.

Poll Question Support Oppose Unsure
Require public disclosure of AI safety-test results 89% 6% 5%
Give federal government power to block risky AI releases 81% 12% 7%

Source: Verasight national survey, June 18–19, 2026, n=1,690 adults.

These findings come at a pivotal moment. AI is moving from lab experiments to daily tools embedded in Windows 11, Microsoft 365, and Azure. Copilot, Recall, and other AI-driven features are reaching hundreds of millions of PCs. Yet there is still no binding federal safety framework for the models powering them. The poll suggests voters want that to change.

What This Means for You

The impact of potential federal AI-safety regulations would cascade across every group that touches Windows—from casual users to enterprise admins and developers. Here’s how the poll’s key demands might reshape your day-to-day experience.

For Home Users and Everyday Windows Customers

If Washington turns these poll numbers into law, you’ll likely see more information about the AI features arriving on your device. Imagine opening Windows Update and seeing not just a feature list for the next cumulative update, but a “Safety Test Summary” link for the latest Copilot model, detailing what risks were checked and how they were mitigated. The same could apply to apps in the Microsoft Store that embed AI capabilities.

Transparency could become the norm. Today, when you opt into Windows Insider builds that test early AI features, you rely on Microsoft’s word that the system has been vetted. A disclosure mandate would add an enforceable layer of documentation—much like nutritional labels forced food companies to reveal what was inside the box. For everyday users, that means more confidence that someone has kicked the tires before forcing an AI change you never asked for.

For IT Administrators and Enterprise Security Teams

The 81 percent figure backing federal power to block risky AI releases has direct enterprise implications. If regulators can halt a model that poses unacceptable security or privacy risks, the pressure shifts to IT teams to ensure their organization’s AI stack stays on the right side of a moving compliance line.

Practical fallout could include:
- Stricter AI inventory requirements: You’ll need to know exactly which models are running in your environment—whether through Copilot in Edge, a third-party plugin, or a line-of-business app.
- Group Policy and MDM controls: Microsoft may accelerate the rollout of enterprise policies that let you block certain AI capabilities or enforce a minimum acceptable safety-test disclosure level before a feature is whitelisted.
- Shadow AI risk: A federal block on a popular AI model could strand employees who adopted it without central approval. The poll’s results suggest that broad public support for blocking risky AI makes it likelier than ever that such a scenario will occur.
- Vendor management: Contracts with SaaS providers will need to spell out who bears liability if a blocked model slips through, turning AI safety into a standard risk-management checkbox.

For Developers and ISVs Building on Windows and Azure

Developers will feel the compliance burden most directly. If disclosure becomes law, your release pipeline must include not just functional tests but also documentation of safety evaluations—red-teaming exercises, bias audits, and robustness checks. That’s a significant shift from today’s voluntary best practices.

But there’s an upside: uniform disclosure requirements can level the playing field. If every AI model must carry a safety “label,” developers can compare models more easily and make better-informed decisions about which APIs to adopt. The poll’s overwhelming support for transparency suggests that users will reward platforms and tools that make safety information easy to find and understand.

How We Got Here: The Long Road to AI Regulation

Public demand for AI guardrails didn’t appear overnight. It grew alongside the technology’s rapid creep into daily life. Here’s a timeline of the major markers that set the stage for this poll:

  • Late 2022: ChatGPT launch triggers a global conversation about AI’s potential and perils.
  • October 2023: President Biden signs an Executive Order on Safe, Secure, and Trustworthy AI, requiring companies to share safety-test results with the government for the largest models. But the order lacks direct enforcement for consumer-facing transparency.
  • 2024: The European Union finalizes the AI Act, creating a risk-based regulatory framework with transparency obligations and prohibited practices. U.S. states, led by Colorado, begin passing their own AI laws, creating a patchwork of rules.
  • 2025: Microsoft announces voluntary commitments to share safety information for some models, but critics note the disclosures are selective and not independently verifiable.
  • Early 2026: Congress holds hearings on comprehensive AI legislation, but bills stall amid partisan disagreements over scope and enforcement. Meanwhile, Microsoft pushes ahead with deep AI integration in Windows 11 24H2 and 25H2, including the controversial Recall feature and pervasive Copilot assistants.

Against this backdrop, the Verasight poll is a temperature check on public sentiment. It confirms that while lawmakers dither, Americans have made up their minds. The message to Washington—and to Redmond—is clear: proceed with AI, but don’t skip the safety briefing.

What to Do Now

Federal legislation is rarely swift, but the poll’s numbers could accelerate the pace. Here are concrete steps for Windows users, IT pros, and developers to prepare.

Home Users

  1. Stay informed about AI updates: In Windows Update > Advanced options, enable notifications for new features. When a feature update mentions AI copilots or model changes, look for accompanying safety documentation—even if it isn’t yet mandatory.
  2. Use Windows Feedback Hub: Your voice matters. If you want to see Microsoft proactively share safety-test results, submit feedback. Overwhelming user demand prompted the company to rework Recall’s privacy controls; safety transparency could be next.
  3. Check your AI privacy settings: Go to Settings > Privacy & security > Diagnostics & feedback, and in copilot-enabled apps review what data is shared. Awareness of data flows is the first step toward understanding model risks.

IT Administrators

  1. Audit AI model usage: Use tools like Microsoft 365 compliance center, Azure AI Content Safety, and third-party inventory solutions to map every AI model touching your network. Document the vendor, version, and known safety certifications.
  2. Leverage existing Group Policies: Microsoft already provides policies for managing Copilot (e.g., “Turn off Copilot in Windows”). Review these and prepare to enforce stricter AI controls if a model is flagged as risky.
  3. Engage with Microsoft’s security baselines: The Security Compliance Toolkit often includes recommended settings for AI features. Check for updates that incorporate safety-test disclosure requirements.
  4. Plan for worst-case scenarios: If a model you rely on is suddenly blocked by a federal regulator, what’s your fallback? Identify alternative models or temporary rollback procedures now.

Developers

  1. Start documenting safety tests today: Even before a law passes, adopt voluntary transparency. Publish a safety-test summary with every model release. Use standardized frameworks like the NIST AI Risk Management Framework or the AI Verify framework.
  2. Build compliance into your CI/CD pipeline: Automate safety checks and generate a report artifact for each release. This will save time when disclosure becomes mandatory.
  3. Monitor policy developments: Subscribe to notifications from the National Institute of Standards and Technology (NIST) and the Federal Trade Commission. Agency rulemaking often precedes legislation.

Looking Ahead: Will Washington Finally Act?

The Verasight poll hands lawmakers a bipartisan mandate: nearly nine in ten voters want something done about AI transparency and safety. While Congress rarely moves fast, the approaching midterm elections could provide a decisive nudge. Candidates in both parties may discover that AI safety resonates as a kitchen-table issue, not an abstract tech debate.

For Microsoft, the poll is both a warning and an opportunity. The company has poured billions into embedding generative AI across its ecosystem. A federal disclosure requirement would mean more paperwork, but it could also insulate Microsoft from future backlash. Proactively publishing comprehensive safety reports for Copilot and Windows AI features—before a law forces the issue—would demonstrate leadership and build trust. Rumors within Microsoft’s engineering circles suggest that a “model transparency hub” is already in early design, though no timeline has been confirmed.

The bigger question is whether a federal blocking power would ever be used. The 81 percent support sends a signal that the public wants a safety valve. Yet actually pulling that valve—halting a major AI release—would be politically and economically seismic. The first test could come sooner than expected: if a Copilot update or third-party model integrated into Windows raises serious safety flags, the call for intervention will move from poll question to front-page news.

For now, the Verasight survey has done something rare: it has made AI regulation feel like a concrete voter expectation rather than a distant policy fight. The next step is up to Washington—and to the companies shipping AI into every PC.