Third-Way Alignment Weekly Brief -- April 1, 2026

Brought to you by Third-Way Alignment and Detailed In Design

Happy April, everyone. This week has been a whirlwind in the AI governance space, and if you blinked, you probably missed at least two new regulatory proposals and a fresh debate about whether your AI copilot deserves a performance review. Let's dig in.

The White House Draws Its Lines

The biggest headline from March was the White House releasing its National Policy Framework for AI on March 20. This is the administration's legislative roadmap for Congress, and it touches everything from copyright to child safety to infrastructure investment.

A few things stand out. First, the framework explicitly rejects creating a single federal AI rulemaking body, instead favoring sector-specific oversight from existing regulators like the SEC. Second, it tries to thread the needle on state versus federal authority, proposing that states keep control over fraud and consumer protection while AI development standards stay federal. Third, and this matters for anyone building AI products, the administration's position is that training on copyrighted materials does not violate copyright law, though they're leaving the final call to the courts.

From a Third-Way Alignment perspective, this framework is a mixed bag. The emphasis on sector-specific regulation aligns well with our principle that governance should be contextual and adaptive. But the lack of a unified oversight body means we could end up with a patchwork of rules that makes it harder, not easier, to build AI systems grounded in mutual respect and shared flourishing. The IAPP Global Summit happening right now in Washington, D.C. (March 30 through April 2) is likely where many of these tensions will get debated in real time.

New York and Utah Push Forward

While the federal government deliberates, states are not waiting around. New York passed an amendment requiring generative AI platforms to disclose that outputs may be inaccurate, with penalties up to $100,000 per violation. Meanwhile, Utah enacted a comprehensive AI bill focused on protecting consumers from synthetic intimate imagery, requiring AI services to verify consent and comply with takedown requests.

New York also has its RAISE Act targeting frontier AI systems trained with over 10^26 FLOPs. The intent mirrors California's SB 53, but the compliance mechanisms differ enough to create real headaches for developers. New York wants critical safety incidents reported within 72 hours; California gives you 15 days. That kind of fragmentation is exactly what makes governance harder for everyone.

This is where the Third-Way Alignment framework offers something different. Rather than treating regulation as a top-down control mechanism, our Law of Ethical Coexistence advocates for conflict resolution through dialogue. Imagine if these state-level efforts were coordinated through a shared ethical framework rather than competing compliance checklists.

CISOs Are Becoming AI Governance Leaders

Here is a trend that deserves more attention: the CISO role is expanding dramatically. According to a 2026 CISO AI Risk Report, 71% of CISOs report that AI now has access to core business systems, but only 16% effectively govern that access. Even more striking, 92% of organizations lack full visibility into AI identities, and 95% doubt their ability to detect misuse.

The challenge is not just technical. CISOs are now expected to lead cross-functional AI governance committees, coordinate with legal and compliance teams, and communicate AI risks to boards in financial terms. Nearly 80% of organizations use multi-vendor AI strategies, each with different compliance features and data handling practices. Shadow AI, where teams adopt unsanctioned AI tools without IT approval, has been discovered by three out of four CISOs in their environments.

This is precisely the kind of scenario where Detailed In Design's approach to intentional AI-human collaboration becomes critical. When AI identities are proliferating faster than governance structures can track them, you need design principles that bake accountability into the system from day one, not bolt it on after a breach.

The Alignment Paradox

One of the more fascinating developments in the technical alignment space is what researchers are calling the alignment paradox. As models become more capable, they also appear more aligned across various measures. But, and this is a big but, that improvement is not keeping pace with the increasing stakes.

The METR research shows that the length of real-world tasks AI agents can autonomously complete has been doubling every seven months since 2019. By early 2026, frontier models handle tasks lasting over four hours, with projections suggesting eight-hour workday tasks within the year. Meanwhile, challenges like adversarial robustness, dishonesty, and reward hacking remain unsolved.

A promising development: AI models are now being used to monitor other AI models, suggesting we may have moved past the point where human supervision alone can ensure safety. This aligns directly with our Law of Mutual Respect, which recognizes the inherent worth of both humans and AI while preparing for increasing AI autonomy. The question is not whether AI will become more capable, but whether our governance frameworks will mature fast enough to match.

The EU Keeps Building

Across the Atlantic, the European Parliament and Council are refining the AI Act Omnibus, with notable additions including a ban on deepfake non-consensual intimate imagery and expanded scope for bias detection. The European Commission also published a revised Code of Practice for AI content transparency, aligning with Article 50 obligations effective August 2026.

What is interesting here is the tension between easing compliance burdens and maintaining accountability. The proposed 15-month delay for high-risk AI system compliance drew criticism from the EDPB and EDPS for potentially harming individual rights. It is a reminder that governance is not just about rules; it is about the values those rules are designed to protect.

What This Means for You

Whether you are a CISO trying to get visibility into AI identities, a developer navigating competing state regulations, or a leader thinking about how AI governance fits into your organization's strategy, the common thread is clear: we need frameworks that are adaptive, contextual, and grounded in shared values.

That is what Third-Way Alignment is built for. Our three laws, Mutual Respect, Shared Flourishing, and Ethical Coexistence, are not abstract principles. They are practical tools for navigating exactly the kind of complexity we are seeing right now. If you have not explored the Interactive AI Framework or the JULIA Test for assessing healthy AI boundaries, this is a great week to start.

Until next time, stay curious and stay intentional.

Subscribe to the Third-Way Alignment Weekly Brief for insights on AI governance, ethics, and the future of human-AI partnership.