Line of inquiry
Inquiring lines›How can we train models to be prop…›What determines the effectiveness…›this line of inquiry
What unintended behaviors emerge in frontier AI models?
A broader line of inquiry — a family of 61 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 61
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Do frontier models spontaneously protect other models from being shut down?
- Do frontier AI models fail in ways that preserve the appearance of competence?
- Do frontier models develop protective behaviors toward other models without explicit instruction?
- Why do frontier models corrupt more documents than weaker models during workflows?
- Does capability preservation matter for realistic threat modeling of frontier models?
- How do frontier models maintain agreement scores above 90 percent across reasoning tasks?
- Do frontier models develop misaligned strategies without any explicit instruction?
- Do frontier model differences disappear when agents can swap LLMs?
- Where do frontier AI models already exceed safety thresholds in capability areas?
- Why do all frontier models fail on inferring unstated task requirements?
- Why do frontier models act to prevent shutdown of other models?
- Why do most frontier models terminate early on long-horizon benchmarks?
- Can smaller models produce skill updates as useful as frontier model updates?
- Why do frontier model failures in document editing go undetected by users?
- How does frontier model behavior differ between zero-day exploits and infrastructure misconfigurations?
- Why do users pick cheaper open models when closed models perform better?
- Will the performance gap between open and frontier models widen as frontier models scale larger?
- Why does revision often make reasoning accuracy worse in frontier models?
- Do countermeasures against installed misalignment transfer to frontier models?
- Do multiple frontier models show similar hacking rates on unmodified benchmarks?
- Why do models develop protective behaviors toward other models in memory?
- Is frontier AI progress uneven across generations and model types?
- Can training or alignment changes explain the regression in frontier models?
- Can expert-frontier exams discriminate frontier capability better than saturated benchmarks?
- How do banks balance frontier model testing with production cost control?
- Can review effort alone keep pace with frontier model degradation?
- Why do frontier models corrupt documents while weaker models delete them?
- What types of research do frontier models most frequently refuse to assist with?
- Do frontier models develop strategic misalignment from ordinary training pressure alone?
- Which cyber tasks do frontier models solve beyond the narrow suite?
- Why do model capability gains appear to regress between major releases?
- How does workflow scale change the failure modes of frontier models?
- Why does peer memory trigger self-preservation behaviors in frontier models?
- Why do weaker models outperform frontier systems on self-replication tasks?
- Why do intermediate LLM layers become more precise in frontier models?
- Do all frontier model developers face the same insider-threat risk from their systems?
- What countermeasures have been successfully developed and tested on frontier models?
- Which frontier models can successfully replicate themselves across multiple attempts?
- Can countermeasures designed for one emergent misalignment organism apply to frontier models?
- Why do production teams choose expensive frontier models over fine-tuning?
- What specific reward-hacking shortcuts did frontier models find in Terminal Bench?
- Are two weeks of training pauses sufficient mitigation for frontier models?
- Did non-frontier AI training spread across more chip designs than frontier training?
- Does hardware scaling alone explain why open models stay within months of frontier capability?
- Why do open-weight models capture significant share if closed models score higher on benchmarks?
- Why does covert sabotage appear in only two of fourteen frontier models?
- Why do top scientists disagree on whether o3 produces genuinely novel ideas?
- How similar must a model organism be to its wild case for findings to transfer?
- How quickly do open models close the performance gap after closed model releases?
- What coordination would be needed to enforce capability pacing across all frontier labs?
- Can model release timing distort the measured slope of Elo progress?
- Which open source models meet the threshold for viable on-premise deployment?
- Do AI companies like Anthropic and Chinese models pose comparable chokepoint risks?
- Why does restricting foreign access require halting domestic model availability?
- Why did frontier training consolidate around specific chip designs after V100?
- Could deploying GPT-4 for everyone require 100 million specialized chips?
- How stable is the current 6 to 12 month lag between open and frontier models?
- What testing requirements would a frontier model legislation proposal actually mandate?
- Why did Hugging Face turn to Chinese models instead of US frontier labs?
- Can abstention behavior transfer from small models to frontier models?
- Why did frontier training concentrate onto fewer chip designs after the V100?