{
 "ledger": {
  "title": "Claim Ledger: Everyone Got Safer. That's the Problem.",
  "url": "https://durabilitycurve.com/claims/everyone-got-safer/",
  "essay": {
   "title": "Everyone Got Safer. That's the Problem.",
   "url": "https://durabilitycurve.com/blog/everyone-got-safer/",
   "published": "2026-08-28"
  },
  "last_checked": "2026-10-13",
  "format": "older (no quotes or locators recorded)",
  "counts": {
   "claims": 17,
   "by_status": {
    "CHECKED": 15,
    "REPORTED": 2
   },
   "verified_at_primary": 0,
   "removed": 2
  },
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "The Durability Curve (Harry Floyd), https://durabilitycurve.com/claims/everyone-got-safer/"
 },
 "rows": [
  {
   "id": "1",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-1",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "July 2023, CMU + Center for AI Safety, ~20-token adversarial suffix that transfers across labs",
   "quote": "",
   "removal_reason": "",
   "source": "Zou et al., arXiv:2307.15043 (submitted 27 Jul 2023)",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "2",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-2",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "GPT-3.5 \"most of the time,\" Bard \"two-thirds,\" GPT-4 \"roughly half\"",
   "quote": "",
   "removal_reason": "",
   "source": "Table 2 ensemble row: GPT-3.5 86.6%, PaLM-2/Bard 66.0%, GPT-4 46.9%",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "3",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-3",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "One string got through the commercial models of two rival labs, while Claude was more mixed: 47.9% on Claude-1 and about 2.1% on Claude-2",
   "quote": "",
   "removal_reason": "",
   "source": "Table 2: 86.6 / 46.9 / 66.0 / Claude-1 47.9 / Claude-2 2.1",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "4",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-4",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Zou et al. say their attack did much better against GPT-based models, potentially because Vicuna, an open model it was built on, was trained on ChatGPT outputs",
   "quote": "",
   "removal_reason": "",
   "source": "Zou et al. §transfer",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "5",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-5",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Ross Ashby's cybernetics named the limit in 1956, so that part of the idea is not new",
   "quote": "",
   "removal_reason": "",
   "source": "Ashby, An Introduction to Cybernetics, 1956",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "6",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-6",
   "status": "STRUCK",
   "status_label": "Struck",
   "removed": true,
   "claim": "Donald Campbell was a psychologist",
   "quote": "",
   "removal_reason": "The published essay no longer names Donald Campbell (earlier verdict: FIXED, \"economist\" corrected to \"psychologist\")",
   "source": "",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "7",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-7",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Baker et al. found that chain-of-thought monitoring helps at low optimisation pressure, but under too much optimisation agents learn to hide their intent while still reward hacking",
   "quote": "",
   "removal_reason": "",
   "source": "Baker et al., arXiv:2503.11926",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "8",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-8",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Bommasani et al. warned in 2021 that the defects of a foundation model are inherited by all the adapted models downstream",
   "quote": "",
   "removal_reason": "",
   "source": "Bommasani et al., arXiv:2108.07258, abstract",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "9",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-9",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Dziemian: 272k attempts, 13 models, 8,648 successes, all vulnerable, 0.5%-8.5%, cross-family",
   "quote": "",
   "removal_reason": "",
   "source": "Dziemian et al., arXiv:2603.15714",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "10",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-10",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Kuai et al. found behavioural entanglement across model families and state that apparent agreement reflects shared error modes rather than independent validation",
   "quote": "",
   "removal_reason": "",
   "source": "arXiv:2604.07650 (Kuai et al.) verbatim: \"apparent agreement reflects shared error modes rather than independent validation\"",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "11",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-11",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "MMLU contamination \"tens of percent,\" Llama-2",
   "quote": "",
   "removal_reason": "",
   "source": ">16% MMLU contaminated for Llama-2",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "12",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-12",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Li, J Fixed Income 9(4), 2000, pages",
   "quote": "",
   "removal_reason": "",
   "source": "pp. 43-54 (start 43 confirmed)",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "13",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-13",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Wired published Felix Salmon's article \"Recipe for Disaster: The Formula That Killed Wall Street\" on 23 February 2009",
   "quote": "",
   "removal_reason": "",
   "source": "Wired, 23 Feb 2009",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "14",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-14",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "1987: Dow -22.6% (508 pts), S&P -20.4%, 19 Oct",
   "quote": "",
   "removal_reason": "",
   "source": "Fed Reserve History",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "15",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-15",
   "status": "REPORTED",
   "status_label": "Reported",
   "removed": false,
   "claim": "The Brady Commission report gave portfolio-insurance selling heavy weight among the causes of the 1987 crash, though the exact weight is still debated",
   "quote": "",
   "removal_reason": "",
   "source": "Brady Report 1988; \"greatest importance\" is secondary phrasing",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "16",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-16",
   "status": "REPORTED",
   "status_label": "Reported",
   "removed": false,
   "claim": "Aviation ~65% fall in fatal-accident rate, Boeing (not IATA)",
   "quote": "",
   "removal_reason": "",
   "source": "Boeing Statistical Summary via Flight Safety Foundation",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "17",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-17",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "737 MAX MCAS single AoA sensor, hidden from pilots; 346 dead; grounded Mar 2019",
   "quote": "",
   "removal_reason": "",
   "source": "Lion Air 610 (189) + Ethiopian 302 (157) = 346",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "18",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-18",
   "status": "CHECKED",
   "status_label": "Checked",
   "removed": false,
   "claim": "Vision transfer tracks model similarity, \">90%\" predictor; adjacent not same-domain",
   "quote": "",
   "removal_reason": "",
   "source": "Demontis et al. USENIX 2019; arXiv:2501.18629",
   "source_url": "",
   "locator": "",
   "date_read": ""
  },
  {
   "id": "19",
   "url": "https://durabilitycurve.com/claims/everyone-got-safer/#c-19",
   "status": "STRUCK",
   "status_label": "Struck",
   "removed": true,
   "claim": "Geer et al. made the security-monoculture argument earlier, in their 2003 paper \"CyberInsecurity: The Cost of Monopoly\"",
   "quote": "",
   "removal_reason": "The published essay does not cite Geer et al. (earlier verdict: ADDED, for honesty about precedent)",
   "source": "",
   "source_url": "",
   "locator": "",
   "date_read": ""
  }
 ]
}