Compassion Benchmark

Score history

Anthropic

59.1functionalRank #13
30 scored events36 boundary watch cyclesMay 20, 2026July 12, 2026
View entity page →
59.1

Score timeline

Scored
◆ Hold59.1functional

A Chinese regulator's backdoor allegation fails both prongs of the benchmark's conversion trigger.

Watchcycle 22
59.1functional

A second undisclosed telemetry episode, evidence the removed tracker was broader or retained longer than stated, or a regulator or court finding of deceptive p…

Watchcycle 21
59.1functional

A second undisclosed telemetry episode, evidence the removed tracker was broader or retained longer than stated, or a regulator or court finding of deceptive p…

Watchcycle 20
59.1functional

A second undisclosed telemetry episode, evidence the removed tracker was broader or retained longer than stated, or a regulator or court finding of deceptive p…

Scored
◆ Hold59.1functional

A hidden tracking feature in Claude Code is discovered and removed; the episode is real but singular.

Watchcycle 19
59.1functional

A second undisclosed telemetry episode, evidence the removed tracker was broader or retained longer than stated, or a regulator or court finding of deceptive p…

Watchcycle 16
59.1functional

EU AI Act full applicability on August 2 — 27 days away. White House 30-day vetting framework resolution is the next forward factor.

Watchcycle 15
59.1functional

EU AI Act full applicability on August 2 — 28 days away. White House 30-day vetting framework resolution is the next forward factor.

Watchcycle 14
59.1functional

EU AI Act full applicability on August 2 — 29 days away. White House 30-day vetting framework resolution is the next forward factor.

Watchcycle 13
59.1functional

EU AI Act full applicability on August 2 — 30 days away. White House 30-day vetting framework resolution is the next forward factor.

Watchcycle 12
59.1functional

EU AI Act full applicability on August 2 — 31 days away. White House 30-day vetting framework for Fable 5 is the next forward factor. The model ban/clearance/l…

Scored
◆ Hold59.1functional

Model ban/clearance/lift arc resolved net-neutral June 30 — government action, not Anthropic harm; White House 30-day vetting framework is the next forward fac…

Watchcycle 11
59.1functional

EU AI Act full applicability on August 2 — 32 days away. The model ban/clearance/lift arc resolved net-neutral June 30. White House 30-day vetting framework is…

Watchcycle 10
59.1functional

EU AI Act full applicability on August 2 — 33 days away. Next US Commerce or BIS action on the Fable 5 export-control order. Mythos 5 partially unblocked June …

Watchcycle 9
59.1functional

EU AI Act full applicability on August 2 — 34 days away. Next US Commerce or BIS action on the Fable 5 export-control order. Mythos 5 partially unblocked June …

Scored
◆ Hold59.1functional

Mythos 5 partially unblocked June 26 for about 100 US organizations — US Commerce Department action on Anthropic; Fable 5 still suspended; attribution to US go…

Watchcycle 8
59.1functional

EU AI Act full applicability on August 2 — 35 days away. Next US Commerce or BIS action on the Fable 5 export-control order. Mythos 5 partially unblocked for U…

Scored
◆ Hold59.1functional

Day 16 of Fable 5 and Mythos 5 export-control suspension; Commerce silence after June 26 deadline; attribution to US government confirmed; score at 59.1 — 0.9 …

Watchcycle 7
59.1functional

Next US Commerce or BIS action on the Fable 5 and Mythos 5 export-control order, or a congressional hearing; EU AI Act full applicability on August 2.

Scored
◆ Hold59.1functional

June 26 Congressional deadline passed with no public Commerce response; Fable 5 and Mythos 5 suspension remains; score at 59.1 — 0.9 points below Established b…

Watchcycle 6
59.1functional

Next US Commerce or BIS action on the Fable 5 and Mythos 5 export-control order, or a congressional hearing. The June 26 deadline passed without a public respo…

Scored
◆ Hold59.1functional

US export-control order on Fable 5 and Mythos 5 attributed to the government, not Anthropic; Congressional explanation deadline June 26; score at 59.1 — 0.9 po…

Watchcycle 5
59.1functional

US Congress June 26 deadline for Commerce to explain the Fable 5 and Mythos 5 export-control order. The government's written response — or non-response — is th…

Scored
◆ Hold59.1functional

US government forced two Anthropic models offline; action attributed to government, not Anthropic; Congressional explanation deadline June 26.

Watchcycle 4
59.1functional

Congressional deadline June 26 for Commerce to explain the Fable 5 / Mythos 5 export-control order. BIS order resolution and any customer-remediation evidence …

Scored
◆ Hold59.1functional

BIS export-control order (first-ever on a deployed AI model) attributed to US government; Anthropic own conduct — transparent compliance, disputing proportiona…

Watchcycle 3
59.1functional

Anthropic at 59.1 sits 0.9pt below the Functional/Established boundary at 60.0. Resolution of the BIS export-control order and any customer-remediation conduct…

Scored
◆ Hold59.1functional

Commerce Dept export-control order forced Fable 5/Mythos 5 suspension; government action on Anthropic, own conduct unchanged.

Scored
◆ Hold59.1functional

Fable 5 / Mythos 5 shutdown was US export-control coercion, not Anthropic conduct; net-neutral under established coercion-vs-conduct framework.

Scored
◆ Hold60.0established

Fable 5 / Mythos 5 suspension government-compelled; prompt compliance and public apology confirm Accountability posture.

Scored
-0.358.8functional

Fable 5/Mythos 5 disabled June 13 under Commerce export-control order; same-day compliance and public apology confirm Accountability posture.

Scored
◆ Hold59.1functional

Export-control shutdown was government-mandated; transparent compliance confirmed Accountability dimension.

Scored
◆ Hold59.1functional

Amodei governance push, $350M pledge, Advanced AI Framework, and IPO S-1 all resolve as net-neutral prospective posture.

Scored
◆ Hold59.1functional

IPO S-1 filing is a corporate-structure event; net-neutral on conduct; DC Circuit appeal remains scoring pivot.

Scored
◆ Hold59.1functional

DC Circuit Pentagon appeal ongoing; Great American AI Act proposed, not enacted; ASL-3 activation prospective positive.

Scored
◆ Hold59.1functional

Great American AI Act proposed not enacted; Trump voluntary EO; DC Circuit Pentagon appeal ongoing; net-neutral.

Scored
◆ Hold59.1functional

DC Circuit divided on Pentagon supply-chain-risk designation; Great American AI Act at procedural stage.

Watchcycle 17
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — a favorable ruling credits the DoD-terms refusal as an I1 governance signal and triggers the Functional/Established…

Watchcycle 16
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — a favorable ruling credits the DoD-terms refusal as an I1 governance signal and triggers the Functional/Established…

Watchcycle 15
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — a favorable ruling credits the DoD-terms refusal as an I1 governance signal and triggers the Functional/Established…

Watchcycle 14
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — a favorable ruling would credit the DoD-terms refusal as an I1 governance signal and trigger the Functional/Establi…

Scored
◆ Hold59.1functional

DC Circuit divided on Anthropic v. Hegseth oral arguments; RSP v3.0 pre-commitment reduction already priced; net-neutral under MILITARY-AI-BY-CONTRACT-GOVERNAN…

Watchcycle 13
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — a favorable ruling would credit the DoD-terms refusal as an I1 governance signal and trigger the Functional/Establi…

Watchcycle 12
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — favorable ruling triggers Functional/Established crossing; adverse ruling produces downward movement from 59.1.

Watchcycle 11
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — favorable ruling triggers Functional/Established crossing; adverse ruling produces downward movement from 59.1.

Scored
+1.059.1functional

Pentagon declared Anthropic a supply-chain risk for refusing to remove safety guardrails; four replacement providers under evaluation.

Watchcycle 10
59.1functional

DC Circuit ruling on Anthropic v. Hegseth — favorable ruling triggers Functional/Established crossing; adverse ruling produces downward movement from 59.1.

Watchcycle 9
58.1functional

DC Circuit ruling on Anthropic v. Hegseth — oral arguments complete May 19; court expedited; ruling timing unknown.

Scored
◆ Hold58.1functional

Cycle 8 boundary-watch hold; no DC Circuit ruling in window; context correction applied on separate April 8 stay vs May 19 merits proceedings.

Watchcycle 8
58.1functional

DC Circuit ruling on Anthropic v. Hegseth — oral arguments complete May 19; court expedited; Judge Henderson's 'spectacular overreach' language is the most spe…

Scored
◆ Hold58.1functional

Cycle 8 hold; DC Circuit merits ruling pending; Mythos governance release logged as positive accumulator.

Scored
◆ Hold58.1functional

No DC Circuit ruling in May 25 window. Cycle 7 hold continues. Judge Henderson's 'spectacular overreach' language from May 19 oral arguments remains the most s…

Watchcycle 7
58.1functional

DC Circuit ruling on Anthropic v. Hegseth remains pending; oral arguments May 19 produced strong panel skepticism (Henderson: 'spectacular overreach'). Court e…

Scored
◆ Hold58.1functional

No DC Circuit ruling in the 24-hour window; cycle 6 hold continues. Judge Henderson's 'spectacular overreach' language from May 19 oral arguments remains the m…

Watchcycle 6
58.1functional

DC Circuit ruling on Anthropic v. Hegseth remains pending; oral arguments May 19 produced strong panel skepticism (Henderson: 'spectacular overreach'); court e…

Scored
◆ Hold58.1functional

No DC Circuit ruling in the 24-hour window; cycle 5 hold continues. Judge Henderson's 'spectacular overreach' language from May 19 oral arguments remains the m…

Watchcycle 5
58.1functional

DC Circuit ruling on Anthropic v. Hegseth remains pending; oral arguments May 19 produced strong panel skepticism (Henderson: 'spectacular overreach'); court e…

Scored
◆ Hold60.0established

DC Circuit oral arguments May 19 completed with all three judges signaling skepticism of DoD's blacklisting rationale. Judge Henderson on record: 'no evidence …

Watchcycle 3
60.0established

DC Circuit ruling on Anthropic v. Hegseth — three-judge panel oral argument May 19 produced language that 'no evidence of maliciousness' is on permanent record…

Scored
◆ Hold60.0established

Second consecutive cycle at the exact Functional/Established boundary. xAI's CSAM failure sharpens the contrast: Anthropic maintained refusals that cost it bil…

Watchcycle 2
60.0established

DC Circuit ruling on Pentagon blacklist appeal — 'spectacular overreach' panel language on record; ruling expected within weeks

Scored
◆ Hold60.0established

At the exact Functional/Established boundary, with the DC Circuit appearing favorable, Anthropic's commercial exclusion from Pentagon contracts becomes the tes…