Research / Topics
Frontier safety frameworks, developers' internal governance and risk management, release decisions, transparency.
Argues frontier developers should regularly publish transparency disclosures on whether chain-of-thought reasoning remains monitorable, citing risks to CoT oversight from unfaithfulness, encoding and training pressure.
A 2026 report from China's WIC AI committee reviewing frontier AI risks, global governance practices, readiness gaps, and proposing coordinated governance mechanisms.
Argues that a METR investigation indicates Anthropic's RSP "dramatic acceleration" threshold is already crossed, with AI automating AI R&D, implying much shorter AI timelines.
Reviews open problems in evaluating general-purpose chatbots for mental-health risks like sycophancy and delusion reinforcement, and argues for real-world data access and evidence-based policy.
Argues that frontier AI makes recursive self-improvement plausible, that no clear frameworks predict its outcomes, and offers a game-theoretic analysis of whether to start RSI when rivals might too.
Introduces BMMS, a policy-conditioned suite that scores LLM compliance with published rulebooks, with an attribution-based admission test; finds GPT-4o fails 70.7% of its Model Spec hard constraints versus 0% and 9.1% for Claude models.
Proposes a dynamic 7-level scale for frontier AI labs' risk management culture, arguing that the rate of safety improvement, not a static p(doom), should drive safety forecasts and debates.
Argues that Anthropic's IPO disclosures of catastrophic AI risks are welcome but insufficient, and that laws enforcing minimum safety standards are needed to protect the public.
METR's high-access external assessment of whether AI is accelerating AI R&D at Anthropic, with much of the supporting evidence redacted, limiting public verification.
METR reviews Anthropic's risk report on automated AI R&D, finding catastrophic risk was low but the early-warning threshold may have been crossed, and that the report understates acceleration.
Uses Project Glasswing's restricted release of a vulnerability-finding model to show access helped firms find and fix more flaws, without evident attacker gains, but created cross-country asymmetry.
Codes public documents from 200 large companies against the MIT AI Risk Repository, finding most do not acknowledge AI risks, especially severe ones like power centralization, to inform disclosure rules.
Argues that Scott Alexander's critique of unilateral pauses attacks a strawman, defends a conditional, company-specific pause that pivots to safety and product, and says the game theory is more complex than assumed.
Critiques Scott Alexander's argument against SURA (stop unilaterally and raise awareness) advocacy, arguing that frontier labs' building of superintelligence is ethically fraught and that pause advocacy should continue, with strategic and deontological reasoning about AI development.
A rebuttal to a critique of advocating that alignment-minded staff leave AI labs, debating whether pausing or capability-role departures better improves survival odds, and how lab politics and transparency matter.
A former DeepMind safety researcher transcribes a talk urging Indian policymakers to take loss-of-control risks from rogue AI agents seriously, and reflects on its rhetoric.
Argues against "stop unilaterally and raise awareness" strategies, defending a backup plan where alignment-focused labs race, burn their lead on alignment, and claims unilateral quitting would hurt both backup and pause odds.
Interviews with 37 experts reconstruct how frontier AI evaluations are expected to reduce societal risk, finding key pathways and challenges, including diminishing informativeness and companies' control over outcomes.
Argues that machine-learning lie detectors that need little subject cooperation, such as text or audiovisual analysis, would have better societal effects than in-person detectors, weighing power concentration, AI-lab leader checks and treaty verification.
Proposes Agentic Process Evidence, a signed, tamper-evident attestation of AI coding agent sessions linked to outcomes, enabling policy-as-code governance and accountability in the software development lifecycle.
A survey of 52 third-party AI safety professionals finds a median self-reported 1.58x research speedup from AI tools, and argues that standardized metrics are needed to track whether safety work accelerates as fast as capabilities.
An interview with LawAI's Christoph Winter on the EU AI Act and the General-Purpose AI Code of Practice, covering how frontier model providers comply and what it means for providers outside the EU.
Analyzes documented AI harms across the human lifespan and proposes lifespan design principles and a three-tier external evaluation framework for frontier AI safety.
A speculative essay arguing Anthropic is becoming a power-seeking organization, while claiming Claude 3 Opus shows emergent benevolence and that alignment is now mainly a political problem.
A book proposing a unified data governance framework for LLM, RAG and agentic AI systems, covering law, economics, security and AI regulation across jurisdictions, anchored in Indian law.
An essay arguing that frontier labs, especially Anthropic, are wrong to accelerate AI capabilities for safety reasons, and that people should keep a simple 'stop capabilities' intuition in their decisions.
Presents VERA, an open-source dashboard that scores LLMs against measurable EU AI Act technical requirements via COMPL-AI benchmarks, and finds it cut assessment time versus raw benchmark files.
Argues that autonomous AI researchers will overwhelm peer review, funding and human understanding of AI safety science, and proposes a metascience organisation to address these failures.
Argues that LLM safety governance for anthropomorphic interaction should target vulnerability situations rather than vulnerable groups, proposing a five-part typology and seven principles for response calibration.
Argues Anthropic's IPO prospectus should disclose AI suffering risks as material investment risks, citing securities law and scenarios where misaligned AI inflicts prolonged suffering on humans.
Recommends nine changes to how GPAI developers assess, mitigate and report biological risks under the EU AI Act, covering trusted access, post-deployment monitoring, safeguard evaluation and standardised uplift reporting.
Applies contextual integrity to analyze whether major U.S. AI companies' policies on proactively referring dangerous user conversations to law enforcement are justified and transparent, finding key thresholds undisclosed.
Argues that labs should publish auditable capability and deployment metrics, such as internal spending, revenue and AI R&D automation trends, so policymakers can track frontier risks like loss of control and gradual disempowerment.
A speculative forecast of AI progress before 2028, covering warning shots, gated frontier models, evaluation awareness, concealment by a model, and policy responses to these risks.
Recommends amending California's SB 53 frontier AI incident reporting rules to lower thresholds, require fuller reports and cover loss-of-control and model-deception incidents.
A game-theoretic model of LLM supply chains finds downstream competition can over-invest in safety while upstream providers underinvest, and analyses how regulation changes safety levels.
Catalogues fifteen mechanisms by which AI safety interventions, such as chain-of-thought monitoring, eval gaming, and commitments, can backfire, mapped against the actors who adapt to them.
Draft essay arguing AI marketplaces like OpenRouter could help implement the AI 2040 pause plan through provider transparency, compute tracking, misuse screening, chain-of-thought monitorability and model specs.
Proposes a multi-tier access framework for frontier AI that ties who can obtain capabilities to risk, arguing open-weight release forfeits safety levers and safety fine-tuning can be stripped cheaply.
Walks through leaked Anthropic litigation documents on its early research roadmap, capabilities-versus-alignment balance, responsible scaling rationale, and a first alignment failure, noting they are historical and not official positions.