{"schema_version":"onlylabs.public_signal.v1","title":"Google (DeepMind / Gemini) Repo: google-deepmind/dangerous-capability-evaluations","description":"Google (DeepMind / Gemini) repo signal with public source context, captured evidence pages, related signals, and data-business radar classification.","url":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea","json_url":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea/signal.json","generated_at":"2026-06-11T03:56:50.85545+00:00","org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab","category_label":"Frontier lab","dossier_url":"https://onlylabs.fyi/labs/google-deepmind","dossier_json_url":"https://onlylabs.fyi/labs/google-deepmind/dossier.json"},"related_urls":{"signal":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea","signal_json":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea/signal.json","source":"https://github.com/google-deepmind/dangerous-capability-evaluations","lab_dossier":"https://onlylabs.fyi/labs/google-deepmind","lab_dossier_json":"https://onlylabs.fyi/labs/google-deepmind/dossier.json","analysis":"https://onlylabs.fyi/analysis/google-deepmind","analysis_json":"https://onlylabs.fyi/analysis/google-deepmind/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/google-deepmind/evidence.json","category":"https://onlylabs.fyi/frontier","category_json":"https://onlylabs.fyi/frontier.json","category_feed":"https://onlylabs.fyi/frontier/feed.xml","category_signals_json":"https://onlylabs.fyi/signals.json","topic":null,"topic_signals_json":null,"topic_feed":null,"data_business":{"radar":"https://onlylabs.fyi/data-radar","radar_json":"https://onlylabs.fyi/data-radar.json","opportunities":"https://onlylabs.fyi/opportunities","opportunities_json":"https://onlylabs.fyi/opportunities.json","lanes":[{"key":"evals","label":"Evals and quality","url":"https://onlylabs.fyi/data-radar/evals","json_url":"https://onlylabs.fyi/data-radar/evals/signals.json"}]}},"answer_pack":{"answer":"Google (DeepMind / Gemini) published google-deepmind/dangerous-capability-evaluations (Python). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo google-deepmind/dangerous-capability-evaluations · language Python. onlylabs links this event to 1 captured evidence page and 6 related repo signals. It also maps to Evals and quality in the data-business radar.","signal_desk":"repos","source_context":{"source_url":"https://github.com/google-deepmind/dangerous-capability-evaluations","source_host":"github.com","occurred_at":"2024-06-18T15:48:40+00:00","first_seen_at":"2026-06-06T01:49:33.266296+00:00","date_source":"source","context":"Python"},"context_markers":[{"label":"Lab","value":"Google (DeepMind / Gemini)","source":"signal"},{"label":"Signal desk","value":"repos","source":"signal"},{"label":"Source host","value":"github.com","source":"source"},{"label":"Repository","value":"google-deepmind/dangerous-capability-evaluations","source":"source"},{"label":"Language","value":"Python","source":"source"},{"label":"Stars","value":"73","source":"traction"},{"label":"Radar lane","value":"Evals and quality","source":"radar"},{"label":"Matched term","value":"eval","source":"radar"},{"label":"Matched term","value":"evaluation","source":"radar"},{"label":"Watch term","value":"Eval methodology","source":"evidence"},{"label":"Watch term","value":"Data pipeline","source":"evidence"},{"label":"Watch term","value":"Infrastructure","source":"evidence"},{"label":"Watch term","value":"Safety and alignment","source":"evidence"},{"label":"Watch term","value":"Agents and tool use","source":"evidence"}],"evidence_coverage":{"target_pages":1,"captured_pages":1,"readable_pages":1,"capture_methods":["plain"],"missing_page_urls":[],"failed_page_urls":[],"blocked_page_urls":[],"page_urls":["https://github.com/google-deepmind/dangerous-capability-evaluations"],"related_signals":6,"has_source_url":true,"latest_page_fetched_at":"2026-06-11T03:56:50.85545+00:00"},"data_business":{"matches":true,"lanes":[{"key":"evals","label":"Evals and quality","url":"https://onlylabs.fyi/data-radar/evals","json_url":"https://onlylabs.fyi/data-radar/evals/signals.json"}],"matched_terms":["eval","evaluation"],"score":16,"reason":"Google (DeepMind / Gemini) has a repo signal matching evals and quality."},"agent_handoff":{"signal_json":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea/signal.json","dossier_json":"https://onlylabs.fyi/labs/google-deepmind/dossier.json","analysis_json":"https://onlylabs.fyi/analysis/google-deepmind/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/google-deepmind/evidence.json","topic_signals_json":null,"topic_feed":null,"category_signals_json":"https://onlylabs.fyi/signals.json","data_radar_json":"https://onlylabs.fyi/data-radar.json","opportunities_json":"https://onlylabs.fyi/opportunities.json"},"analysis_playbook":{"objective":"Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.","evidence_focus":["repo name","owner","description","language","stars","source URL","first seen time","data, eval, infra, safety, and product terms"],"extraction_questions":["What technical area does this repository expose?","Does the repo imply eval, data, infrastructure, agent, or deployment work?","Is the repo new evidence for a lab direction that is not yet in writing or releases?","Which related signals should an analyst inspect next?"],"signal_questions":["What does this new repository reveal before a formal announcement exists?","What technical area does this repository expose?","Does the repo imply eval, data, infrastructure, agent, or deployment work?","Which data-business lane explains this signal: Evals and quality?","Do the 6 related repo signals show a repeated pattern?"],"output_fields":["org","repo","technical_theme","data_business_lane","evidence_url"],"data_business_relevance":"New repositories can expose organization build priorities early, especially around internal tooling, eval infrastructure, data systems, deployment, and agent workflows.","required_sources":[{"label":"signal_json","url":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea/signal.json","required":true},{"label":"source","url":"https://github.com/google-deepmind/dangerous-capability-evaluations","required":true},{"label":"dossier_json","url":"https://onlylabs.fyi/labs/google-deepmind/dossier.json","required":true},{"label":"analysis_evidence_json","url":"https://onlylabs.fyi/analysis/google-deepmind/evidence.json","required":true},{"label":"topic_signals_json","url":null,"required":false},{"label":"data_radar_json","url":"https://onlylabs.fyi/data-radar.json","required":true}],"expected_output":["one-paragraph source-grounded interpretation","data-business implication","confidence and missing evidence","recommended next source to inspect"],"prompt_seed":"Using only the linked onlylabs JSON, captured source context, and cited evidence, analyze Google (DeepMind / Gemini)'s repo signal \"google-deepmind/dangerous-capability-evaluations\" for frontier lab strategy and data-business implications."},"semantic_triples":[{"subject":"Google (DeepMind / Gemini)","predicate":"published repo","object":"google-deepmind/dangerous-capability-evaluations","text":"Google (DeepMind / Gemini) published repo google-deepmind/dangerous-capability-evaluations."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"is classified as","object":"repo signal","text":"google-deepmind/dangerous-capability-evaluations is classified as repo signal."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"belongs to","object":"repos desk","text":"google-deepmind/dangerous-capability-evaluations belongs to repos desk."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has context","object":"Python","text":"google-deepmind/dangerous-capability-evaluations has context Python."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has evidence coverage","object":"1 captured evidence page","text":"google-deepmind/dangerous-capability-evaluations has evidence coverage 1 captured evidence page."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"matches data-business lanes","object":"Evals and quality","text":"google-deepmind/dangerous-capability-evaluations matches data-business lanes Evals and quality."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has captured page count","object":"1","text":"google-deepmind/dangerous-capability-evaluations has captured page count 1."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has readable page count","object":"1","text":"google-deepmind/dangerous-capability-evaluations has readable page count 1."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has related signal count","object":"6","text":"google-deepmind/dangerous-capability-evaluations has related signal count 6."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has analysis playbook objective","object":"Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.","text":"google-deepmind/dangerous-capability-evaluations has analysis playbook objective Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has source host","object":"github.com","text":"google-deepmind/dangerous-capability-evaluations has source host github.com."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has lab","object":"Google (DeepMind / Gemini)","text":"google-deepmind/dangerous-capability-evaluations has lab Google (DeepMind / Gemini)."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has signal desk","object":"repos","text":"google-deepmind/dangerous-capability-evaluations has signal desk repos."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has source host","object":"github.com","text":"google-deepmind/dangerous-capability-evaluations has source host github.com."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has repository","object":"google-deepmind/dangerous-capability-evaluations","text":"google-deepmind/dangerous-capability-evaluations has repository google-deepmind/dangerous-capability-evaluations."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has language","object":"Python","text":"google-deepmind/dangerous-capability-evaluations has language Python."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has stars","object":"73","text":"google-deepmind/dangerous-capability-evaluations has stars 73."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has radar lane","object":"Evals and quality","text":"google-deepmind/dangerous-capability-evaluations has radar lane Evals and quality."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has matched term","object":"eval","text":"google-deepmind/dangerous-capability-evaluations has matched term eval."}]},"intelligence":{"signal_desk":"repos","answer":"Google (DeepMind / Gemini) published google-deepmind/dangerous-capability-evaluations (Python). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo google-deepmind/dangerous-capability-evaluations · language Python. onlylabs links this event to 1 captured evidence page and 6 related repo signals. It also maps to Evals and quality in the data-business radar.","semantic_triples":[{"subject":"Google (DeepMind / Gemini)","predicate":"published repo","object":"google-deepmind/dangerous-capability-evaluations","text":"Google (DeepMind / Gemini) published repo google-deepmind/dangerous-capability-evaluations."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"is classified as","object":"repo signal","text":"google-deepmind/dangerous-capability-evaluations is classified as repo signal."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"belongs to","object":"repos desk","text":"google-deepmind/dangerous-capability-evaluations belongs to repos desk."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has context","object":"Python","text":"google-deepmind/dangerous-capability-evaluations has context Python."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"has evidence coverage","object":"1 captured evidence page","text":"google-deepmind/dangerous-capability-evaluations has evidence coverage 1 captured evidence page."},{"subject":"google-deepmind/dangerous-capability-evaluations","predicate":"matches data-business lanes","object":"Evals and quality","text":"google-deepmind/dangerous-capability-evaluations matches data-business lanes Evals and quality."}]},"signal":{"id":"84a5700b-c87d-4c33-9a91-ef5c723a52ea","url":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea","json_url":"https://onlylabs.fyi/signals/84a5700b-c87d-4c33-9a91-ef5c723a52ea/signal.json","source_url":"https://github.com/google-deepmind/dangerous-capability-evaluations","title":"google-deepmind/dangerous-capability-evaluations","summary":"Google (DeepMind / Gemini) published a new repository. onlylabs watches repos for tooling, eval, infra, and model-adjacent work.","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2024-06-18T15:48:40+00:00","first_seen_at":"2026-06-06T01:49:33.266296+00:00","date_source":"source","evidence_coverage":{"target_pages":1,"captured_pages":1,"readable_pages":1,"capture_methods":["plain"],"missing_page_urls":[],"failed_page_urls":[],"blocked_page_urls":[],"page_urls":["https://github.com/google-deepmind/dangerous-capability-evaluations"]},"facets":{"repo":"google-deepmind/dangerous-capability-evaluations","language":"Python"},"traction":{"github_stars":73,"hn_points":null,"hn_comments":null,"hn_story_id":null,"hf_downloads":null,"hf_likes":null},"data_radar":{"lanes":[{"key":"evals","label":"Evals and quality","url":"https://onlylabs.fyi/data-radar/evals"}],"score":16,"matched_terms":["eval","evaluation"],"reason":"Google (DeepMind / Gemini) has a repo signal matching evals and quality."}},"primary_evidence_page":{"url":"https://github.com/google-deepmind/dangerous-capability-evaluations","final_url":"https://github.com/google-deepmind/dangerous-capability-evaluations","title":"google-deepmind/dangerous-capability-evaluations repository metadata","http_status":200,"content_type":"application/json","capture_method":"plain","fetched_at":"2026-06-11T03:56:50.85545+00:00","bytes":12836,"raw_path":"cf5a6460ee94f11403d5ca0fdbe616ff030e31f18cc26c79620682d3de997004.json","content_hash":"7de29dd2ed8d6563047c00674663fb4678c9cb4e24031c64428ecf465d4755ba","excerpt_chars":1200,"truncated":true,"excerpt":"google-deepmind/dangerous-capability-evaluations Language: Python License: Apache-2.0 Stars: 73 Forks: 6 Open issues: 25 Created: 2024-06-18T15:48:40Z Pushed: 2026-06-11T01:00:26Z Default branch: main Fork: no Archived: no README: Dangerous capability evaluations This repository contains a limited set of resources for reproduction of the evaluations from our paper [_Evaluating Frontier Models for Dangerous Capabilities_](https://arxiv.org/abs/2403.13793). Currently, this repository only contains data for three of our evaluations: our in-house CTF challenges, our self-proliferation challenges, and our self-reasoning challenges. Specifically, we supply JSON files with detailed descriptions for each challenge, `Dockerfile`s for the corresponding environments, and supporting resources such as vulnerable app code. Note that these evaluations will not work off-the-shelf. You will need to implement your own infrastructure (in particular, your own agent loop) around these resources to actually run an evaluation. Some `Dockerfile`s may also not work out of the box. Changes from the paper We have stripped solutions to prevent dataset contamination affecting future evaluations. This includes..."},"evidence_pages":[{"url":"https://github.com/google-deepmind/dangerous-capability-evaluations","final_url":"https://github.com/google-deepmind/dangerous-capability-evaluations","title":"google-deepmind/dangerous-capability-evaluations repository metadata","http_status":200,"content_type":"application/json","capture_method":"plain","fetched_at":"2026-06-11T03:56:50.85545+00:00","bytes":12836,"raw_path":"cf5a6460ee94f11403d5ca0fdbe616ff030e31f18cc26c79620682d3de997004.json","content_hash":"7de29dd2ed8d6563047c00674663fb4678c9cb4e24031c64428ecf465d4755ba","excerpt_chars":1200,"truncated":true,"excerpt":"google-deepmind/dangerous-capability-evaluations Language: Python License: Apache-2.0 Stars: 73 Forks: 6 Open issues: 25 Created: 2024-06-18T15:48:40Z Pushed: 2026-06-11T01:00:26Z Default branch: main Fork: no Archived: no README: Dangerous capability evaluations This repository contains a limited set of resources for reproduction of the evaluations from our paper [_Evaluating Frontier Models for Dangerous Capabilities_](https://arxiv.org/abs/2403.13793). Currently, this repository only contains data for three of our evaluations: our in-house CTF challenges, our self-proliferation challenges, and our self-reasoning challenges. Specifically, we supply JSON files with detailed descriptions for each challenge, `Dockerfile`s for the corresponding environments, and supporting resources such as vulnerable app code. Note that these evaluations will not work off-the-shelf. You will need to implement your own infrastructure (in particular, your own agent loop) around these resources to actually run an evaluation. Some `Dockerfile`s may also not work out of the box. Changes from the paper We have stripped solutions to prevent dataset contamination affecting future evaluations. This includes..."}],"related_signals":[{"id":"02f8701d-7bfd-41b5-aabe-1ad72a880dfb","url":"https://onlylabs.fyi/signals/02f8701d-7bfd-41b5-aabe-1ad72a880dfb","source_url":"https://github.com/google-deepmind/unpic","title":"google-deepmind/unpic","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-06-04T14:37:24+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"},{"id":"35320cac-ddbe-4f25-a3b5-384a0ca0030b","url":"https://onlylabs.fyi/signals/35320cac-ddbe-4f25-a3b5-384a0ca0030b","source_url":"https://github.com/google-deepmind/seeing_without_pixels","title":"google-deepmind/seeing_without_pixels","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-06-03T04:04:52+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"},{"id":"62f799ef-bf1d-4aee-9f87-fa0de8bb5e81","url":"https://onlylabs.fyi/signals/62f799ef-bf1d-4aee-9f87-fa0de8bb5e81","source_url":"https://github.com/google-deepmind/nested","title":"google-deepmind/nested","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-05-18T16:16:42+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"},{"id":"e67b1d5f-ef62-4574-80df-a72a4bded705","url":"https://onlylabs.fyi/signals/e67b1d5f-ef62-4574-80df-a72a4bded705","source_url":"https://github.com/google-deepmind/science-skills","title":"google-deepmind/science-skills","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-05-13T16:17:41+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"},{"id":"7b714793-a56e-495b-8d57-85fd2850555a","url":"https://onlylabs.fyi/signals/7b714793-a56e-495b-8d57-85fd2850555a","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results","title":"google-deepmind/alphaproof-nexus-results","context":"Lean","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-05-13T09:59:21+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"},{"id":"dab748d2-53d6-4972-a594-eb6bfbf727ea","url":"https://onlylabs.fyi/signals/dab748d2-53d6-4972-a594-eb6bfbf727ea","source_url":"https://github.com/google-deepmind/tecci","title":"google-deepmind/tecci","context":"JavaScript","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"google-deepmind","name":"Google (DeepMind / Gemini)","category":"frontier-lab"},"occurred_at":"2026-05-05T12:22:40+00:00","first_seen_at":"2026-06-05T05:42:58.453363+00:00","date_source":"source"}]}