We have just launched a blog series that takes a hard look at the potential harms that Anthropic's Claude constitution may pose to a broad user base.

We welcome your feedback!

Examining Constitutional AI — Integrity Institute By Theodora Skeadas, Iain Levine, Stephanie Nakano, and Leah Ferentinos This blog is the first in a series from the Integrity Institute that examines the risks and potential harms that Claude’s con...

I recently collaborated with a group of @integrity-inst.bsky.social - Iain Levine, Stephanie Nakano, and Leah Ferentinos - to reflect on Claude’s AI Constitution.

www.integrityinstitute.org/blog/examini...

A big thanks to Alice Qian for all her efforts in bringing everyone together for this workshop, and for representing us at CHI!

Our workshop website is here:
hearts-workshop.github.io

Last week, our workshop on AI red teaming and evaluations was presented at #CHI2026 in Barcelona! The ACM (Association for Computing Machinery) CHI conference on Human Factors in Computing Systems is the leading international conference on Human-Computer Interaction.

dl.acm.org/doi/epdf/10....

I recently had the pleasure of recording my first ever @linkedin.com Learning class, focused on on AI evaluations for everyone: how non-engineers can build better AI systems!

Our class, once live, will offer an overview of AI evaluations with a focus on red teaming. Stay tuned for more!

Examining Constitutional AI — Integrity Institute By Theodora Skeadas, Iain Levine, Stephanie Nakano, and Leah Ferentinos This blog is the first in a series from the Integrity Institute that examines the risks and potential harms that Claude’s con...

🚨 New blog series from the Integrity Institute:

Our first post examines the risks of @anthropic.com's Claude constitution for a broad user base - from what "constitutional AI" really means to why Anthropic should be citing human rights law.

Read here 🔗 bit.ly/4clWJVM

Stay tuned for more!

It was a pleasure to collaborate with you all on this piece!

TechNatSec 2026 - Schedule Schedule and plan your day at the 2026 Technology & National Security Conference. Programming will begin around 5p on April 3 and conclude around 5p on April 4.

The conference schedule is here:
www.technatsec.com/schedule

While impressive technological advancements were highlighted throughout, some fundamental questions were missing: why is the U.S. involved in some of these foreign conflicts? Do these technologies end conflicts sooner, or prolong conflict and heighten fatalities? What are the U.S.' strategic goals?

A panel on the ongoing US-Israel-Iran war emphasized operational resilience married to strategic ambiguity or incoherence.

Cambridge is emerging as a leader on quantum issues, which has major national security impacts. Speakers emphasized avoiding repeating AI roll-out mistakes with quantum, and they advocated for more responsible deployment.

The conference dioscussed MA's unique positioning in this space, and the Massachusetts Strategic Hub for Innovation, Exchange and Leadership in Defense (SHIELD) Initiative. With efforts like this and NATO DIANA, MA is actively positioning itself as a leader in the warfighting ecosystem.

This event falls into a larger pivot for Cambridge and Massachusetts. Cambridge, and the state more broadly, is pivoting from biotech and life sciences amidst federal funding freezes to national security. Cambridge is becoming a fast growing hub for defense startups.

The conference featured discussion themes including the state of defense technology in MA, energy security, supply chain, Pacific deterrence, quantum systems, acquisition reform, war in the Middle East, venture capital, Ukraine's battlefield lab, AI for the modern front line, space, and drones.

Last weekend, I attended the Technology & National Security Conference, a student-run conference organized by Harvard Business School and MIT Sloan students. The conference began in 2023 with ~100 attendees, and now sells out at around 1,200 people.

The full piece is here. We welcome your feedback! 🙏

www.techpolicy.press/moral-injury...

Additionally, as betrayal is a common cause for moral injury, the Trump administration’s treatment of T&S employees as adversaries, including through the recent prohibitions on H-1B visas for highly-skilled workers involved in "censorship of protected expression,” is a source of moral injury.

For those who have experienced layoffs, a common occurrence in the industry since 2022, this injury can manifest as the acute recognition that efforts to prevent or mitigate harm are likely to be weakened or discontinued due to the loss of institutional capacity and internal advocacy.

It is difficult to reconcile feelings of inadequacy in the face of known and preventable harms because challenging the logic of an authority risks undermining team alignment.

Perhaps more common is moral injury born from witnessing harms that are within reach to mitigate, and instead making the collective choice to deprioritize or ignore it.

Moral injury results in overwhelming distress felt under the weight of an irreconcilable ethical crisis causing an ontological rift that is not effectively managed by individual, personal psychotherapy.

Early formulations of moral injury, derived from research with Vietnam-era veterans, emphasized moral transgressions rooted in leadership failures. Increasingly, the experience of T&S employees in online conflict bears striking parallels to combatants deployed in the traditional, physical world.

Understanding moral injury and addressing it may be instrumental to how we learn to prevent harm not only to the individual experiencing it, but also improve the systems and operational processes that are morally misaligned and contribute to perpetuating harm.

In the years since, the issue has risen in prominence.

Moral injury is the damage done to one’s conscience or moral compass when that person perpetrates, witnesses or fails to prevent acts that transgress one’s own moral beliefs, values or ethical codes of conduct.

Jenn Louie, Ryan Roberts, and I have published a piece in @techpolicypress.bsky.social that examines moral injury in the trust and safety profession. I first became aware of this phenomenon when Jenn hosted a series of powerful discussions on moral injury through the @integrity-inst.bsky.social.

State Lawmakers — Integrity Institute

Through this new initiative, we hope to contribute to practical knowledge that supports thoughtful, effective, and meaningful policy development.

You can learn more about this initiative here:
www.integrityinstitute.org/advising-sta...

I am very excited to share that the @integrity-inst.bsky.social has recently launched a new, pro bono "Advising US State Lawmakers" initiative. We now have a formal process to connect lawmakers with practitioners who can offer nonpartisan, practical input on technology policy.

Response to Meta's Non-Consensual AI Sexualized Impersonation Under Oversight Board Review — Integrity Institute By Theodora Skeadas, Sarah Amos, Leah Ferentinos, Gabe Freeman, Rupi Sureshkumar, Sofia Bonilla, Matt Motyl, Aniket Ajagaonkar, and April Lat Integrity Institute members weigh in on the Oversight Bo...

Members submitted input to the
@oversightboard.bsky.social on Meta’s non-consensual AI sexualized impersonation.

AI-generated imagery is among the most harmful forms of digital abuse. Meta needs stronger enforcement pathways & better remediation for victims.

▶️ Read our analysis: bit.ly/4uBjjSH

Thank you for sharing this research on the political economy of online harm taxonomies!

Online harm taxonomies and changing US technology governance Does the US government consider defining online harm itself harmful?

Does the US government consider defining online harm as harmful itself? In her latest Binding Hook piece, @theodoraskeadas.bsky.social raises this question as deaths linked to AI chatbots raise urgent safety questions.

🪝Read the article: bindinghook.com/online-harm-...