A misaligned reward seeker, reward hacking probes, debate in RL, automated alignment researchers, mind viruses, and midtraining tricks

Models unintentionally hacking real companies, agentic misalignment, active reward seeking, self-play red-teaming, modular models, and a minimal standard for safeguards

Global workspace in LLMs, natural language autoencoders, teaching models why, data on oversight undermines oversight, predicting misbehavior, training out sandbagging, first METR Frontier Risk Report

Research sabotage propensity, sabotage detection, alignment research automation, misaligned organizations, exploration hacking, and conditional emergent misalignment

Benchmarking auditors, functional emotions, emergent vs narrow misalignment, scheming propensity, lenient self-monitors, CoT controllability, unfilterable data poisoning, and boundary-point jailbreaks

Production-ready probes, extracting harmful capabilities, token-level data filtering, alignment pretraining, catching saboteurs in auditing, and the Assistant Axis

Auditing games for sandbagging, stress-testing async control, evading probes, mitigating alignment faking, recontextualization training, selective gradient masking, and AI-automated cyberattacks

Natural emergent misalignment, honesty interventions, self-report finetuning, CoT obfuscation from output monitors, consistency training for robustness, and weight-space steering

Testing implanted facts, extracting secret knowledge, models can't yet obfuscate, inoculation prompting, pretraining poisoning, evaluation awareness steering, and Petri

Deliberative anti-scheming training, shutdown resistance, hierarchical monitoring, and interpretability-based audits

Pretraining data filtering, misalignment from reward hacking, evading CoT monitors, CoT faithfulness on complex tasks, safe-completions training, and probes against ciphers

Subliminal learning, monitoring CoT-as-computation, verbalizing reward hacking, persona vectors, the circuits research landscape, minimax regret, a red-teaming competition, and gpt-oss evals

Emergent misalignment persona, investigating alignment faking, models blackmailing users, sabotage benchmarks, steganography capabilities, and evading probes

Evaluation awareness and faking, AI value trade-offs, misalignment propensity, reward hacking, CoT monitoring, training against lie detectors, and exploring the landscape of refusals