Rujuta Karekar
AI governance and safety researcher. Mechanical engineer by training. Focused on military AI governance, safety frameworks, and the structural gaps in how frontier AI is constrained.
I studied mechanical engineering and worked as an engineer. That's where I learned to think about how things actually break — not how they're supposed to work on a whiteboard.
The move to AI governance came from noticing that the hardest problems in this space are structural. Who controls deployment. Where oversight degrades. What happens when systems behave outside their design parameters. Engineering doesn't hand you the answers, but it gives you the right instincts for the questions.
Activities
Stop Killer Robots — Youth Network
Meaningful human control over autonomous weapons. UNIDIR poster. UNSG recommendations. Digital colonialism blog. Youth advocacy toolkit. Humanitarian concerns report in progress.
AI Safety Coordination Hub
Co-founder. Building a coordination space for AI safety researchers and practitioners.
Also: education volunteering across several organisations, and structured AI safety coursework (alignment, governance, AGI safety). Details on the CV.
I'm building a body of work at the intersection of engineering thinking and governance research. Interested in collaboration — and in being told where the framing is wrong.
When I'm not doing any of this, I write fiction — mostly romance, mostly for fun.
Research & Writing
Frameworks, posters, policy inputs, and longer analytical writing.
Open questions
What governance mechanisms apply when a company's deployment policy constrains military AI use more than international law does?
Can red lines for AI systems be made concrete and measurable — not just aspirational?
How does loss of control actually unfold — structurally, gradually, through institutional erosion?
Beyond job displacement: what are we losing in literacy, camaraderie, and collective capacity?
Outputs & Presentations
RED30: 30 Indicators for AI Red Lines
Apart Research Technical Governance Hackathon
30 indicators, four severity tiers, grounded in international treaties. Interactive dashboard.
AI Red Lines Tracker Dashboard
Interactive assessment tool
Evaluates AI systems against ethical and legal thresholds.
Meaningful Human Control and AI in Military Use
UNIDIR Conference
Structural requirements for human control over autonomous weapons.
Virtue Ethics Approaches to AI Safety
Tokyo AI Safety Conference 2025
Alignment-as-character vs alignment-as-constraint.
Autonomous Systems as Digital Colonialism
Stop Killer Robots
Autonomous weapons development as a reproduction of colonial power asymmetry.
AI Red Lines Tracker and RED30 Demo
Apart Research Technical Governance Hackathon
Framework and dashboard demonstration.
Youth and Humanitarian Concerns with AI Weapons
Stop Killer Robots (in progress)
Youth perspectives on humanitarian implications of AI weapons development.
Essays & Analysis
Essays coming soon.
In the meanwhile, read here →
Policy Interests
Three areas that organise my research. Each represents a question I consider both important and insufficiently addressed.
Military AI Governance & Autonomous Weapons
Who governs military AI when the companies building the models make deployment decisions that no treaty or procurement regulation accounts for? The civil-military distinction most frameworks rely on has already broken down. I work on this through research at Sentient Future and advocacy with Stop Killer Robots.
AI Safety Frameworks & Red Lines
Governance standards mean nothing if they can't be operationalised. RED30 is my attempt to build concrete, measurable indicators grounded in international law. Separately, I'm researching virtue ethics as an alternative alignment paradigm with Prof Mizumoto.
Loss of Control & Structural Risk
Loss of control is a process, not an event. Institutional drift, compounding oversight failures, gradual erosion of human authority. My engineering training — thinking about cascading failures and system degradation — directly applies. This is the focus of my work at Sentient Future.
Current Focus
Last updated: April 2026
Research
AGI loss-of-control pathways at Sentient Future
Virtue ethics and AI safety with Prof Mizumoto
Developing RED30 into a more robust assessment tool
Gradual disempowerment — Effective Thesis Accelerator
Policy & Advocacy
Youth advocacy toolkit for autonomous weapons — Stop Killer Robots
Humanitarian concerns report on AI weapons (in progress)
Community
AI Safety Coordination Hub — co-founder
Open to
Research fellowships in AI governance and policy
RA roles at think tanks, labs, or policy organisations
Collaboration on military AI, autonomous weapons, or structural AI risk
I'm interested in working with people across governance, safety, and military AI. If there's overlap — or disagreement — reach out. → email