Upcoming AI safety events

Annual workshop on mechanistic interpretability at ICML, focused on advancing understanding of neural networks. Topics include high-level properties of models, using interpretability to identify and fix undesirable behaviors, and expanding to more realistic settings on larger frontier models. Non-archival workshop welcoming work from academia, industry, and independent researchers.

{
  "type_weight": 0.85,
  "source_trust": 0.9,
  "topic_relevance": 0.95,
  "time_proximity": 0.7383647798742139,
  "community_signal": 0.85,
  "speaker_org_signal": 0.8,
  "is_cfp_open": 1,
  "source_count": 2
}

A 4-month full-time research fellowship at Anthropic focused on AI safety and alignment. Fellows work on scalable oversight, adversarial robustness, AI control, model organisms, mechanistic interpretability, AI security, and model welfare. Includes weekly stipend of $3,850 USD, ~$15k/month compute funding, and close mentorship from Anthropic researchers. Over 80% of first cohort fellows produced papers, and over 40% subsequently joined Anthropic full-time.

{
  "type_weight": 0.8,
  "source_trust": 0.95,
  "topic_relevance": 0.95,
  "time_proximity": 0.6880503144654088,
  "community_signal": 0.8,
  "speaker_org_signal": 0.95,
  "is_cfp_open": 1,
  "source_count": 1
}

A 5-day unconference for 100+ researchers focused on theoretical AI alignment. Topics include Singular Learning Theory, Agent Foundations, and Causal Incentives. Third iteration of this gathering. Free to attend with limited travel and accommodation support available. Organized by Iliad, an umbrella organization for applied mathematics research in alignment.

{
  "type_weight": 1,
  "source_trust": 0.85,
  "topic_relevance": 0.95,
  "time_proximity": 0.6176100628930817,
  "community_signal": 0.75,
  "speaker_org_signal": 0.75,
  "is_cfp_open": 1,
  "source_count": 1
}

OpenAI's external research fellowship for rigorous, high-impact research on safety and alignment of advanced AI systems. Fellows work on priority areas including safety evaluation, ethics, robustness, scalable mitigations, privacy-preserving safety methods, agentic oversight, and high-severity misuse domains. Program includes monthly stipend, compute support, API credits, and ongoing mentorship. Fellows work in Berkeley at Constellation or remotely, partnering with OpenAI mentors.

{
  "type_weight": 0.8,
  "source_trust": 0.9,
  "topic_relevance": 0.95,
  "time_proximity": 0.40628930817610065,
  "community_signal": 0.8,
  "speaker_org_signal": 0.95,
  "is_cfp_open": 1,
  "source_count": 1
}

A fully funded, in-person five-month fellowship at Constellation's Berkeley research center. Fellows work on AI safety research projects paired with expert mentors. The program includes travel support, accommodation, and stipend. Strong placement rates at leading AI safety organizations.

{
  "type_weight": 0.8,
  "source_trust": 0.85,
  "topic_relevance": 0.9,
  "time_proximity": 0.40628930817610065,
  "community_signal": 0.7,
  "speaker_org_signal": 0.75,
  "is_cfp_open": 1,
  "source_count": 1
}

A weekend hackathon focused on AI safety research, organized by Apart Research. Part of their series of global AI safety hackathons with emphasis on participation from Global South researchers and practitioners.

{
  "type_weight": 0.65,
  "source_trust": 0.85,
  "topic_relevance": 0.85,
  "time_proximity": 0.8440251572327044,
  "community_signal": 0.6,
  "speaker_org_signal": 0.7,
  "is_cfp_open": 1,
  "source_count": 1
}

A 12-week research fellowship designed to train and support emerging researchers in AI alignment and related fields. Fellows work on technical governance, empirical research, policy & strategy, theory, or compute governance tracks. Top-performing fellows can extend for 6-12 months of additional research. Applications are currently closed but collecting expressions of interest.

{
  "type_weight": 0.8,
  "source_trust": 0.95,
  "topic_relevance": 0.95,
  "time_proximity": 0.9345911949685535,
  "community_signal": 0.85,
  "speaker_org_signal": 0.9,
  "is_cfp_open": 0,
  "source_count": 1
}

Third annual Technical AI Safety Conference. A free, one-day workshop on AI safety covering a broad range of AI safety research topics. Organized by Oxford Martin AI Governance Initiative and Noeon Research. Brings together researchers, practitioners, and policymakers to discuss technical approaches to AI safety.

{
  "type_weight": 1,
  "source_trust": 0.9,
  "topic_relevance": 0.95,
  "time_proximity": 0.8571428571428571,
  "community_signal": 0.8,
  "speaker_org_signal": 0.85,
  "is_cfp_open": 0,
  "source_count": 1
}

ARENA (Alignment Research Engineering Accelerator) is a 4-5 week intensive in-person bootcamp in London focused on AI alignment research engineering. The program covers mechanistic interpretability, reinforcement learning, and other technical alignment topics, with fully funded travel, accommodation, and meals.

{
  "type_weight": 0.8,
  "source_trust": 0.9,
  "topic_relevance": 0.95,
  "time_proximity": 0.969811320754717,
  "community_signal": 0.75,
  "speaker_org_signal": 0.8,
  "is_cfp_open": 0,
  "source_count": 1
}

Intensive 9-week fully-funded research program for 30 fellows advancing careers in AI safety. Focus areas include alignment, interpretability, formal verification, multi-agent safety, AI governance, technical governance (compute governance, model evals, standards), and economics of transformative AI. Fellows receive $10k stipend, housing, up to $10k in compute credits/GPU access, free weekday meals, weekly mentorship from Harvard/MIT/Northeastern researchers, and networking opportunities.

{
  "type_weight": 0.8,
  "source_trust": 0.85,
  "topic_relevance": 0.95,
  "time_proximity": 0.89937106918239,
  "community_signal": 0.8,
  "speaker_org_signal": 0.8,
  "is_cfp_open": 0,
  "source_count": 1
}

3-month fellowship to launch or accelerate impactful careers in American AI governance and policy. Fellows conduct independent research projects under expert mentorship while building professional networks and developing policy expertise. Focus areas include public policy, political science, engineering, economics, biosecurity, cybersecurity, China studies, and risk management. Prioritizes bipartisan engagement, rigorous analysis, and practical policy relevance. $21,000 stipend plus travel support, weekday lunches, and DC office space. US work authorization required.

{
  "type_weight": 0.8,
  "source_trust": 0.9,
  "topic_relevance": 0.9,
  "time_proximity": 0.89937106918239,
  "community_signal": 0.8,
  "speaker_org_signal": 0.85,
  "is_cfp_open": 0,
  "source_count": 1
}

3-month fellowship for conducting independent research on AI governance topics. Fellows receive mentorship from field experts, participate in seminars and Q&A sessions, and build professional networks. Research outputs may include reports, white papers, journal articles, op-eds, or blog posts. £12,000 stipend plus travel support and weekday lunches. Open to candidates from government, academia, industry, or civil society with expertise in policy, political science, computer science, economics, or risk management. Visa sponsorship available.

{
  "type_weight": 0.8,
  "source_trust": 0.9,
  "topic_relevance": 0.9,
  "time_proximity": 0.89937106918239,
  "community_signal": 0.8,
  "speaker_org_signal": 0.85,
  "is_cfp_open": 0,
  "source_count": 1
}

4-month research fellowship with Anthropic mentors focusing on scalable oversight, adversarial robustness, AI control, mechanistic interpretability, and AI security. Fellows receive $3,850/week stipend plus ~$15k/month compute funding. Fellows choose and shape research projects in close collaboration with Anthropic researchers. Over 80% of previous fellows published research.

{
  "type_weight": 0.8,
  "source_trust": 0.95,
  "topic_relevance": 0.95,
  "time_proximity": 0.4857142857142857,
  "community_signal": 0.85,
  "speaker_org_signal": 0.95,
  "is_cfp_open": 0,
  "source_count": 1
}

Three-month fully-funded research fellowship for scholars in economics, law, international relations, and related fields focusing on societal impacts of advanced AI and institutions/policies for effective response. Fellows receive $25,000 stipend, covered travel, daily meals, and access to CAIS expertise and Bay Area network. Emphasizes producing publicly shareable research on AI's impact on economic distribution, corporate accountability, and geopolitical competition.

{
  "type_weight": 0.8,
  "source_trust": 0.9,
  "topic_relevance": 0.85,
  "time_proximity": 0.9345911949685535,
  "community_signal": 0.75,
  "speaker_org_signal": 0.9,
  "is_cfp_open": 0,
  "source_count": 1
}

Second iteration of the Pluralistic Alignment Workshop at ICML 2026, exploring approaches to aligning AI systems with diverse human values and preferences. Part of the main conference workshop track.

{
  "type_weight": 0.85,
  "source_trust": 0.85,
  "topic_relevance": 0.95,
  "time_proximity": 0.7333333333333334,
  "community_signal": 0.8,
  "speaker_org_signal": 0.7,
  "is_cfp_open": 0,
  "source_count": 1
}

A one-day workshop bringing together global leaders in academia and industry to deepen collective understanding of potential risks from Artificial General Intelligence (AGI). Organized by FAR.AI, focusing on alignment research and risk mitigation strategies.

{
  "type_weight": 0.85,
  "source_trust": 0.9,
  "topic_relevance": 0.9,
  "time_proximity": 0.7584905660377359,
  "community_signal": 0.7,
  "speaker_org_signal": 0.85,
  "is_cfp_open": 0,
  "source_count": 1
}

Workshop on secure and verifiable AI development, bringing together researchers, builders, and funders across ML, hardware security, systems, cryptography, and computer security. Focuses on verification techniques for AI safety. Colocated with IEEE Security and Privacy conference. Organized by FAR.AI.

{
  "type_weight": 0.85,
  "source_trust": 0.9,
  "topic_relevance": 0.85,
  "time_proximity": 0.9428571428571428,
  "community_signal": 0.65,
  "speaker_org_signal": 0.85,
  "is_cfp_open": 0,
  "source_count": 1
}

Second Workshop on Technical AI Governance Research at ICML 2026, focusing on technical approaches to AI governance, policy, and regulation. Part of the main conference workshop track.

{
  "type_weight": 0.85,
  "source_trust": 0.85,
  "topic_relevance": 0.9,
  "time_proximity": 0.7383647798742139,
  "community_signal": 0.75,
  "speaker_org_signal": 0.7,
  "is_cfp_open": 0,
  "source_count": 1
}

Workshop at ICML 2026 focused on identifying, diagnosing, and fixing failure modes in agentic AI systems. Covers reproducible triggers for failures, diagnostic tracing methods, and verified repair approaches. Highly relevant to AI safety and robustness.

{
  "type_weight": 0.85,
  "source_trust": 0.85,
  "topic_relevance": 0.9,
  "time_proximity": 0.7383647798742139,
  "community_signal": 0.75,
  "speaker_org_signal": 0.7,
  "is_cfp_open": 0,
  "source_count": 1
}

Second Workshop on Agents in the Wild focusing on safety and security of AI agents deployed in real-world environments. Addresses challenges in ensuring safe and secure operation of autonomous agents. Part of ICML 2026 workshop track.

{
  "type_weight": 0.85,
  "source_trust": 0.85,
  "topic_relevance": 0.9,
  "time_proximity": 0.7333333333333334,
  "community_signal": 0.75,
  "speaker_org_signal": 0.7,
  "is_cfp_open": 0,
  "source_count": 1
}

First global academic programme focused on AI evaluations. 90 hours online (lectures, networking, activities), 20 hours hands-on courses, and 40-hour in-person capstone week in Valencia. 40 fully-funded scholars receive a 15 ECTS Expert Diploma via ValgrAI. Faculty from Cambridge, Stanford, Princeton, EU AI Office, UK AI Safety Institute, FAR AI, Apollo Research, and other leading institutions.

{
  "type_weight": 0.8,
  "source_trust": 0.85,
  "topic_relevance": 0.9,
  "time_proximity": 0,
  "community_signal": 0.75,
  "speaker_org_signal": 0.9,
  "is_cfp_open": 0,
  "source_count": 1
}

Foresight Institute's 40th anniversary event featuring a dedicated AI safety track titled 'Emerging AI Paradigms: Navigating the future of intelligence' powered by Apollo Research. Three-day conference bringing together 300+ frontier academics, builders, and researchers. Includes 30+ expert presenters, technology demonstrations, and expert office hours. While broader in scope than pure AI safety, the dedicated track led by Apollo Research makes it relevant to the community.

{
  "type_weight": 1,
  "source_trust": 0.75,
  "topic_relevance": 0.6,
  "time_proximity": 0.9144654088050315,
  "community_signal": 0.5,
  "speaker_org_signal": 0.75,
  "is_cfp_open": 0,
  "source_count": 1
}