{"id":"W4417101019","doi":"10.70777/si.v2i4.16671","title":"International AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management","year":2025,"lang":"","type":"article","venue":"SuperIntelligence - Robotics - Safety & Alignment","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Key (lock); Risk management; SAFER; Risk assessment; Safeguard; Corporate governance; Risk governance","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","sts","insufficient_payload"],"consensus_categories":["metaepi_narrow"],"category_scores_codex":[0.003977937,0.001388208,0.001322303,0.0007069146,0.001452243,0.00100099,0.004015722,0.0006407808,0.0026646],"category_scores_gemma":[0.0004464759,0.001535504,0.0005437824,0.001555812,0.001008206,0.001180313,0.007950736,0.002242721,0.0002645967],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001952371,"about_ca_system_score_gemma":0.0005244201,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001678956,"about_ca_topic_score_gemma":0.00009254226,"domain_scores_codex":[0.9884675,0.0007138882,0.003545367,0.003414403,0.002157464,0.001701432],"domain_scores_gemma":[0.9944174,0.0004904865,0.0008444645,0.003151747,0.0005027934,0.0005931635],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0004431104,0.0005083493,0.005998868,0.0002778447,0.001414009,0.001258363,0.0007779606,0.4701502,0.00004760301,0.3320492,0.001494022,0.1855805],"study_design_scores_gemma":[0.001654243,0.0003259418,0.01208614,0.001036694,0.0008193096,0.0006527143,0.001011902,0.4330889,0.0008346702,0.01480465,0.531373,0.002311853],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0002391154,0.001197408,0.9278736,0.01265224,0.008797884,0.001640164,0.0001182301,0.000307492,0.04717391],"genre_scores_gemma":[0.4822765,0.04803244,0.4436655,0.002419196,0.0008358224,0.0001538408,0.0002449504,0.0001938775,0.02217783],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.529879,"threshold_uncertainty_score":0.9998868,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.009785592565935903,"score_gpt":0.2868525432628055,"score_spread":0.2770669506968696,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}