Research Teams
AI safety and security research teams.
AI safety and security research teams.
Research Teams
• Three Laws is an agentic AI alignment research collaboration investigating how fundamental principles from biology and economics can inform safer, more aligned AI systems. • Our work centres on homeostasis, multi-objective balancing, sustainability, and universal human values — drawing from nature's time-tested strategies for maintaining equilibrium — to develop benchmarks that elicit and expose dangerous runaway failure modes in current AI approaches. • We also research frameworks that mitigate these risks. We believe that shifting AI design from "maximise forever" toward "maintain a healthy equilibrium" is a crucial and underexplored part of the alignment solution space.
A research collective doing basic research and empirical work on large multi-agent systems. We are interested in threat models like gradual disempowerment, power concentration, and the intelligence curse. We are interested in objects like swarms, agent societies, agent social media, agent economies, agent markets, and hybrid human-AI epistemic environments.
AIXI Labs models AI risk factors and safety mitigations in terms of AIXI variants, and develops the means to translate them to real AI agents. AIXI is the leading mathematical model of artificial superintelligence—the maximum theoretical limit of AI capabilities. This enables rigorous testing of both the risk factors and the safety mitigations.
sociotechnial AI safety and alignment towards superintelligence.
We are a small independent research team working towards wise AGI.
AI strategy, forecasting, and institutional engagement in the Philippines
B-Side Labs is building the science of AI character under pressure. We design evaluations, benchmarks, and lightweight instruments to measure and, eventually, mitigate character instability in frontier models and agentic systems.
Fangcun Leap builds defense systems that protect AI models, LLMs, and autonomous agents from adversarial threats. Our team also participate writing OpenClaws users' safety guidance and regulations in China.
Architecting AI focuses on adversarial red-teaming and alignment stress-testing to map the structural limitations and foundational failure modes of frontier large language models. The lab applies spatial systems thinking to treat LLMs as high-dimensional informational architecture, utilising defensive red-teaming and applied capability testing for deep, entity-level verification.
Parallax is a non-profit research lab developing a second, deeper view of model behaviour. Our mission is to make white-box auditing of agents’ internal beliefs, goals, and plans a core layer of frontier AI evaluations by developing methods and infrastructure that support this depth of analysis at scale. By combining behavioural evidence with evidence drawn from a model’s internal computations, we want to diagnose alignment failures and distinguish the underlying mechanisms that produce them.
To protect human relevance in the age of advanced AI, policy-makers must identify and work to protect key human technical capacities.
Individual Researchers
I research how military and security uses of AI affect human rights, accountability and the balance of power, with a focus on Latin America and other middle-power states. Current work covers autonomous weapons regulation, AI procurement in defense and security (including a registry of public AI contracts in Colombia), algorithmic sovereignty, and the material supply chain of AI (critical minerals, data centers and community rights). I’m interested in collaborating on comparative work across the Global South.
We work on the interaction between AI and society, with questions spanning responsible AI, AI alignment, and the role of incentives in AI environments.
My independent research focuses on AI safety, alignment, and the evaluation of multilingual language models. I investigate reliability gaps in how language models process identity-related information across different languages, scripts, and code-switched contexts. My work aims to develop empirical evaluation methods for identifying model failures and understanding their implications for safe and reliable AI deployment. I am also interested in model behavior, evaluation design, and culturally and linguistically robust AI systems.
Independent AI safety researcher with an ML/NLP background since 2016, most recently the sole AI engineer at a US health-tech startup, building LLM systems in production. Happy to contribute to existing teams and projects, especially where research engineering meets interpretability or theory. Currently exploring: reading formal structure out of transformer weights, i.e. treating a transformer as an automaton whose state is its KV cache and computing its minimal automaton from the weights alone. More broadly I'm drawn to deconfusion-style questions and to interpretability methods that make few assumptions about architecture or data.
Through individual and mentored research, I investigate components of self-awareness in LLMs. These include metacognition, introspection, self-modeling, self-perspective taking, and endogenous preferences/goals. Primary methods are behavior-based evaluations and whitebox and blackbox interpretability.
I focus on anything I think is doable and will move the needle. Mostly focused around steering and representation engineering as pragmatic tools for alignment and improving aI character.
Currently focused on chain of thought monitoring, evaluations, and encoded reasoning
Mycelium builds the technical foundation for AI to consider the welfare of all sentient beings - humans, nonhuman animals, and digital minds. Concretely, we work on technical research & engineering - developing benchmarks, evals, data pipelines and other research experiments to build an evidence base for frontier labs to take seriously the moral patienthood of nonhuman welfare.
We study how AI systems form, maintain, and revise interpretations over time, especially in ambiguous or multi-agent settings. Our work focuses on misunderstanding, overconfidence, premature convergence, perspective shifts, and the mechanisms that help models remain open to corrective evidence. More broadly, we are interested in the dynamics of meaning: how competing interpretations emerge, how prior reasoning trajectories shape future beliefs, and how AI systems can detect when they may be confidently wrong and recover more effectively. The goal is to develop AI systems that are more epistemically flexible, corrigible, and robust under uncertainty.