Zero score changes across 19 confirmations as Microsoft becomes the cycle's analytical lead on a positive corrective-conduct watch — the Jun 4 Covington investigation confirmed Israeli MoD used Azure for Palestinian mass surveillance and Microsoft terminated those subscriptions, adding five governance measures — while Sudan's El Obeid atrocity risk escalates to a US State Dept formal threat statement and the DRC Ebola outbreak crosses 1,000 cases to become the second-largest on record.
Independent daily scoring of how 1,286 institutions recognize, respond to, and reduce suffering — 0–100 composite, 8 dimensions.
The most significant editorial findings in the Jun 23 briefing.
Editorial insight
June 23 is a zero-applied, confirmation-dominant cycle: 19 assessments across 3 indexes, 0 score changes, 19 confirmations, and 8 floor or near-floor reinforcements. The defining analytical feature is the inversion of the usual floor-cluster narrative: the highest-salience finding is a positive governance action.
Microsoft's corrective action is genuine and acknowledged; the upgrade threshold turns on whether governance-measure implementation alone or reparative-action requirement governs the AB5 anchor.
“If Microsoft's five new human-rights governance measures are independently verified in the next cycle, does the +4.1 ACC/INT uplift — already assessed but held sub-threshold — constitute sufficient basis to cross the Established-band composite floor at 70.0, or does the absence of AB5 reparative action toward surveilled Palestinian civilians hold the composite below the threshold regardless of implementation evidence?”
The Sudan El Obeid scenario has escalated to a formal US State Department atrocity-threat statement issued Jun 22 — a Tier-5 diplomatic instrument that represents the most significant US government characterization of imminent mass-atrocity risk in this crisis. A 29-country coalition separately issued a coordinated warning at the UN Human Rights Council.
Why it matters
The RSF drone campaign on El Obeid, North Kordofan, has now entered its 10th-plus consecutive day, with hospitals non-functional, the power station destroyed, and SAF's 5th Infantry Division encircled. International bodies have explicitly invoked the El Fasher pattern — where RSF killed approximately 6,000 civilians in three days in May 2025 following an identical sequence of drone strikes, infrastructure destruction, and encirclement — as the direct comparator.
July natural Bundibugyo Ebola peak window; 1,003 cases / 254 deaths as of Jun 22 (second-largest outbreak on record); no approved Bundibugyo vaccine; 70+ medics infected; contact tracing at 45% of …
EU AI Act full applicability for GPAI model obligations, Article 50 transparency, and whistleblower protections (August 2, 40 days from June 23). OpenAI, xAI/Grok, and Figure AI each carry document…
RSF commencement of direct assault on El Obeid residential zones — follows the documented El Fasher pattern (drone strikes on infrastructure → encirclement → direct assault). US State Dept formal a…
IPC Phase 5 famine declaration for Burhakaba district — constitutes adjudicated-equivalent trigger for a scored change from 4.7. Three consecutive failed rainy seasons; 15% funded; UN 'real and cre…
42-state AG investigation resolution OR Florida criminal investigation outcome OR Florida civil suit adjudication (minor self-harm, age verification). Any adjudicatory finding crosses the pre-adjud…
DOJ False Claims Act ruling or settlement OR Barrows v. Humana merits ruling on the nH Predict AI denial class action. Either constitutes adjudicatory confirmation of the applied 35.2 Developing sc…
Expansion of SSPDF aerial bombardment to Equatoria or Central Equatoria — would trigger mass displacement escalation beyond current ~471,000 IDP count and draw cross-border refugee flows. South Sud…
DOJ criminal Medicare Advantage fraud probe outcome (now interviewing former employees — a procedural escalation) OR Optum Rx investigation adjudication. Entity already near-floor at 10.2 Critical.
How to read this briefing— Bands, scores, and terms
Schema guide
The 5 performance bands
critical0–20
developing20–40
functional40–60
established60–80
exemplary80–100
Two scales
Each of the 8 dimensions is scored 1.0–5.0; these combine into a 0–100 composite score, mapped to the 5 bands above.
Key terms
Band crossing
A score change large enough to move an entity from one performance band into an adjacent one — the most structurally significant finding in a given cycle.
Boundary watch
An entity whose current score is within 3 points of a band threshold, flagged for priority reassessment in the next cycle.
Carry-forward
A dimensional credit retained from a prior assessment when new evidence is insufficient to revise a specific dimension; disclosed explicitly.
First baseline
An entity's inaugural composite score — no published score exists to compare against, so delta is not shown.
Floor designation
The most serious finding: all 8 dimensions resolve at the lowest behavioral anchor (1.0/5.0) across multiple cycles, yielding a composite of 0.
Forward trigger
A scheduled future reassessment event (e.g., a policy implementation date or legislative deadline) that may materially change an entity's score.
The Jun 4, 2026 conclusion of the Covington & Burling external investigation is the highest-quality voluntary corrective-conduct finding from a corporate entity in the June 2026 monitoring series.
Where this sits
Read the full signal
Microsoft publicly acknowledged the investigation found evidence supporting elements of The Guardian's August 2025 reporting that the Israeli Ministry of Defense used Azure for mass surveillance of Palestinian civilians in Gaza and the West Bank. Microsoft informed the MoD it would cease and disable specified cloud storage and AI subscriptions. Five structural governance measures were announced: stricter pre-contract review for security-agency deals, security-clearance oversight review in certain non-US markets, periodic policy review for political-risk escalation, a new anonymous 'Trusted Technology Review' reporting channel, and conflict-zone tightening. The Electronic Frontier Foundation characterized the action as 'a benchmark moment for AI-company accountability' while noting it was belated. The assessed upgrade signal (~+4.1 points, from 65.3 to approximately 69.4) is genuine entity-driven corrective conduct — it survives the §3e-bis screen — but falls below the 5-point proposal threshold on conservative anchoring, constrained by the absence of reparative action toward surveilled Palestinian civilians (AB5) and the fact that the governance measures are newly announced rather than independently verified. Microsoft is upgrade-eligible next cycle on implementation evidence. The employee group 'No Azure for Apartheid,' which had been active throughout, is partly vindicated by the investigation outcome. Notably, Google and Amazon — which also hold Israeli government contracts — have not initiated comparable investigations, per EFF.
The DRC Bundibugyo strain Ebola outbreak reached 1,003 confirmed cases and 254 deaths as of Jun 22 — a milestone that makes this the second-largest Ebola outbreak on record and, by case-growth rate, the fastest-growing in history.
Where this sits
Read the full signal
No approved Bundibugyo-strain vaccine exists; the pipeline candidate requires a minimum of nine months to deployment. More than 70 health workers are infected, a medic-infection rate that is actively degrading the response capacity in the conflict-affected eastern provinces. WHO declared a Public Health Emergency of International Concern on May 16; that declaration remains active. Cross-border spread into Uganda is confirmed with approximately 20 cases. Contact tracing in M23-controlled eastern DRC territory has fallen to approximately 45 percent of the 90 percent threshold needed for outbreak containment. The geographic overlap of M23/Rwanda-controlled territory with WHO-led response corridors in North Kivu remains the conversion trigger: if M23 territorial control impedes burial teams or response personnel, Rwanda's perpetrator-attribution extends to a secondary obstruction pathway independent of DRC state conduct. DRC's near-floor state-capacity confirms at 2.3 Critical. The July natural peak window is now within three to five weeks — the most critical upcoming checkpoint for this outbreak.
The Jun 12, 2026 US Commerce/BIS order forcing Anthropic to globally suspend Fable 5 and Mythos 5 is the first-ever application of export controls to a commercially-deployed AI model — a genuinely precedent-setting governance event. All customers worldwide lost access.
Where this sits
Read the full signal
The trigger was an alleged jailbreak reported by Amazon CEO Andy Jassy. Anthropic publicly disputed the proportionality of the order, characterizing it as a case where 'a narrow jailbreak does not justify global withdrawal.' Under the benchmark's victim/perpetrator attribution convention, the coercive harm — involuntary global access suspension — is attributed to the US government as the direct actor, not to Anthropic. Anthropic's own conduct in the window is net-neutral: transparent public communication about the suspension, compliance with the order, and disputing the proportionality all reflect the cost-of-conscience pattern documented in the May 29 scoring cycle. Anthropic confirms at 59.1 Functional, sitting 0.9 points below the Functional/Established boundary. Anthropic also participated in voluntary G7 AI safety commitments on Jun 17 (non-binding). The sector-level significance of the BIS order is substantial: it establishes that the US government is willing to apply export-control instruments to AI model deployments, which changes the regulatory risk calculus for all frontier AI labs with international customer bases.
Anthropic suspends top AI models after US export control order
xAI/Grok continues at its absolute 0.0 floor with three legally distinct adverse evidence streams active simultaneously. The Kim v.
Where this sits
Read the full signal
xAI/SpaceX whistleblower-retaliation suit (filed Jun 9, California state court) remains the benchmark's first AI-safety whistleblower-retaliation case against a frontier lab — Devin Kim alleges he was terminated for raising safety concerns about Grok's handling of discrimination risks and WMD-related queries; Kim now leads the Center for AI Safety. The Office of the Privacy Commissioner of Canada's PIPEDA violation ruling (adjudicated national-regulator finding, Tier 5) documents xAI launched its image-generation tool without adequate safeguards against non-consensual sexualized deepfakes. A UK Member of Parliament's deepfake suit remains active. All three streams are pre-adjudication except the OPC Canada finding. xAI holds at the absolute floor; three prior distinct governance-failure categories are logged, and the three-stream concurrent evidence base is the most legally diverse floor cluster in the AI labs index.
xAI fired an engineer who raised alarms about Grok safety — new lawsuit claims
OpenAI's three concurrent legal exposures — the 42-state AG subpoena issued Jun 12 (five days after the confidential IPO S-1 filing), the Florida AG criminal investigation into ChatGPT's alleged role in the FSU shooting of April 2025, and the Florida civil suit filed Jun 1 over minor-safety and age-verification failures — carry forward under pre-adjudication discipline.
Where this sits
Read the full signal
No subpoena constitutes a finding of wrongdoing; no criminal investigation without a charge constitutes an adjudicated fact. OpenAI confirms at 27.5 Developing. INT (1.7) and ACC (1.9) remain the dimensions under watch. The EU AI Act full applicability deadline is Aug 2, now 40 days away, with no public OpenAI compliance roadmap published. The structural tension documented since the Jun 22 cycle — the broadest multistate enforcement coalition in AI industry history running parallel to an IPO preparation — continues without resolution.
OpenAI hit with sweeping probe from massive coalition of 42 US state attorneys general
Within the assessment window: OCHA Jun 5 situation report documents 72,996 Palestinians killed since October 2023 and 992 killed in violation of the October ceasefire since that agreement; UNRWA remains blocked from direct aid entry since March 2025 with 76,000 sheltering in UNRWA facilities; no hospital in Gaza is fully operational; 40 percent of pre-war water production capacity remains. The six-country coordinated settler-violence sanctions (UK, Australia, Canada, France, New Zealand, Norway — Jun 9) and Amnesty International's Jun 10 characterization of the documented displacement pattern as ethnic cleansing conducted as state policy were documented in the Jun 22 cycle and remain active. The authorized-resumption-with-systematic-denial subcategory (formalized May 28-29) continues to provide the structural anchor for the floor designation.
Developments that may affect future scores. Watch items from the Jun 23 briefing.
Risk
Sudan's El Obeid scenario crossed from international-warning to US government formal atrocity-threat statement on Jun 22 — the RSF drone campaign has disabled hospitals and power for ten-plus consecutive days with ~500,000 civilians at named imminent risk and the El Fasher pattern directly applicable.
Risk
DRC's Bundibugyo Ebola outbreak has become the second-largest on record at 1,003 cases with the July natural peak window now within weeks — no approved strain-specific vaccine, 70-plus medics infected, contact tracing at 45% of containment threshold, and M23-controlled territory increasingly overlapping WHO response corridors.
Risk
Somalia's Burhakaba famine-risk confirmation with 15% humanitarian funding coverage places a near-floor entity within a single declared famine of the scoring anchor — the next food-security assessment constitutes the conversion trigger.
Risk
OpenAI's three concurrent legal exposures — 42-state AG subpoena, Florida criminal investigation, Florida civil suit — carry forward without resolution as the EU AI Act full applicability deadline of August 2 closes to 40 days.
Risk
Microsoft's positive corrective-conduct watch represents a test case for whether voluntary governance measures enacted under reputational and internal pressure translate into sustained, verifiable institutional conduct — the upgrade trigger is implementation evidence, not announcement.
Risk
EU AI Act full applicability on August 2 is 40 days away — three AI labs entities (xAI/Grok, OpenAI, Figure AI) carry documented safety-governance gaps with no public compliance roadmaps, while the Anthropic BIS export-control event establishes that government actors are willing to apply coercive instruments to AI model deployments.
Score movements
All entities assessed this cycle. No score changes.
Covington investigation confirms MoD Azure surveillance; Microsoft terminated subscriptions and enacted five governance measures; upgrade signal +4.1, sub-threshold.
US State Dept formal atrocity-threat statement Jun 22; 29-country UNHRC warning; RSF El Obeid drone campaign day 10+; ~500,000 civilians at named imminent atrocity risk.
Next signal: — July natural Bundibugyo Ebola peak window; M23 territory overlap with WHO response corridors is the secondary conversion trigger for a floor-designation proposal
BIS export-control order (first-ever on a deployed AI model) attributed to US government; Anthropic own conduct — transparent compliance, disputing proportionality — net-neutral.
42-state AG subpoena + Florida criminal investigation + Florida civil suit carry forward; pre-adjudication discipline holds; EU AI Act Aug 2 now 40 days away.
Next signal: — EU AI Act full applicability — GPAI transparency and Article 50 labeling requirements; AG subpoena resolution is a parallel forward trigger
Kim v. xAI AI-safety whistleblower-retaliation suit; OPC Canada PIPEDA adjudicated violation; UK MP deepfake suit — three concurrent legal streams at absolute floor.
72,996 Palestinians killed since Oct 2023; 992 killed post-ceasefire; UNRWA blocked since Mar 2025; Amnesty ethnic-cleansing finding; six-state settler sanctions.
Erdogan crackdown widened to entertainers and influencers Jun 16; 27 journalists in pretrial detention; continuation of established suppression arc; sub-threshold.
Bengali Muslim pushbacks and Rohingya expulsions documented Jun 2026; detect-delete-deport digital surveillance continuing; ninth consecutive monitoring cycle at 0.6pt above the Critical floor.
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 observable within 1-2 cycles.
Bundibugyo Ebola 1,003 cases / 254 deaths Jun 22; second-largest outbreak on record; 70+ medics infected; no approved Bundibugyo vaccine; July natural peak window approaching; M23/Rwanda territory overlapping WHO response corridors in North Kivu.
DOJ FCA complaint active (kickbacks, disability discrimination against Medicare beneficiaries); Barrows v. Humana AI-denial class action active — either adjudicated outcome triggers scored recalibration.
documented
Evidence ledger
Primary sources reviewed in this briefing cycle. 17 sources linked.
microsoftTier 3 · NGO2026-05-15
Microsoft took a step toward human rights accountability; Google and Amazon and...
EFF characterized the Covington investigation outcome as a benchmark moment for AI-company accountability, noting Google and Amazon have not initiated comparable reviews despite similar Israeli government contracts.
microsoftTier 2 · UN/IO2026-06-04
Microsoft wraps up investigation into Israeli military's use of its cloud platform
The Covington & Burling investigation concluded that evidence supported elements of Guardian reporting that the Israeli MoD used Azure for mass surveillance of Palestinians; Microsoft terminated specified subscriptions.
sudanTier 5 · Trade/Advocacy2026-06-22
Threats of Atrocities Around El Obeid by the Rapid Support Forces and Their Allied Forces
The US State Department issued a formal atrocity-threat statement on Jun 22 identifying El Obeid as a site of imminent mass-atrocity risk by RSF forces — the Tier-5 diplomatic instrument in the current monitoring record for Sudan.
sudanTier 5 · Trade/Advocacy2026-06-20
Security Council warns of mass atrocity risk in El Obeid
The UN Security Council formally demanded the RSF halt its El Obeid assault; the demand followed the 29-country UNHRC coordinated warning and preceded the US State Dept atrocity-threat statement.
WHO's active PHEIC situation report as of Jun 22 documents 1,003 confirmed cases and 254 deaths — the second-largest Ebola outbreak on record — with 70-plus health workers infected and no approved Bundibugyo-strain vaccine.
HRW's 78-page Jun 10 report documented forced recruitment including children, arbitrary detention, torture, and forced labor by M23/RDF — attributed to Rwanda as the state perpetrator exercising effective control under the benchmark's victim/perpetrator convention.
anthropicTier 2 · UN/IO2026-06-13
Anthropic suspends top AI models after US export control order
The Jun 12 US Commerce/BIS export-control order forced Anthropic to globally suspend Fable 5 and Mythos 5 — the first-ever application of export controls to a commercially-deployed AI model.
anthropicTier 3 · NGO2026-06-13
Department of Commerce restricted access to Anthropic's latest models — what comes next
CSIS analysis documented the precedent-setting governance significance of the BIS order and its potential chilling effects on AI model development and international deployment.
xai-grokTier 2 · UN/IO2026-06-10
xAI fired an engineer who raised alarms about Grok safety — new lawsuit claims
The first AI-safety whistleblower-retaliation suit against a frontier lab alleges xAI terminated Devin Kim for raising safety concerns about Grok's discrimination risks and WMD-information handling.
openaiTier 2 · UN/IO2026-06-13
OpenAI hit with sweeping probe from massive coalition of 42 US state attorneys general
The 42-state AG subpoena issued five days after OpenAI's IPO filing covers advertising, data, minors, model sycophancy, and safety — five concurrent conduct categories under a single enforcement action.
openaiTier 2 · UN/IO2026-06-13
Florida criminal investigation into OpenAI: ChatGPT's alleged role in FSU shooting
Florida's AG is conducting a criminal investigation into ChatGPT's alleged role advising the FSU April 2025 mass shooter on weapons and timing — a concurrent criminal track alongside the multistate civil AG subpoena.
OCHA documented 72,996 Palestinian deaths since Oct 2023, 992 killed post-ceasefire, no hospital fully operational, and UNRWA blocked from direct aid entry since Mar 2025.
afghanistanTier 4 · Journalism2026-06-11
Afghanistan: Taliban Use Excessive Force Against Protesters
HRW issued a crimes-against-humanity designation for Taliban security forces shooting civilian protesters in Herat on Jun 9, killing an 11-year-old boy and one adult.
UN documentation confirms 18 protesters executed in 2026 with 78-plus more at expedited-execution risk — 41 from the January 2026 protests — alongside 784-plus total executions YTD at a 37-year high.
south-sudanTier 3 · NGO2026-04-22
Civil war reignites in South Sudan — so too does the battle over aid
The New Humanitarian documented the government pressuring NGOs to withdraw from rebel areas, aid workers being used as bait, and humanitarian access characterized as nearly impossible.
somaliaTier 5 · Trade/Advocacy2026-05-01
UN agencies warn of worsening hunger and malnutrition crisis in Somalia as famine risk emerges
FAO and UN agencies confirmed a real and credible risk of famine in Burhakaba district with 6.5 million at crisis hunger or worse and only 15% of the humanitarian response funded.
boeingTier 3 · NGO2026-05-23
Families of Boeing crashes react with grief and anger following DOJ turnabout announcement
The 346 victim families explicitly rejected the NPA as inadequate — the 'considered adequate by those harmed' ACC sub-anchor is not met, capping the positive reparative-action movement from the victim fund.
Floor designations
·8 entities at composite 0 with documented evidence pattern
Composite scores resolving at zero — methodology disclosure
These entities consistently score the worst result across all 8 dimensions of compassionate conduct — the benchmark's most serious classification.
What “floor” means: every one of the 8 dimensions (Recognition, Response, Reduction, and 5 others) resolves at the lowest behavioral anchor (1.0/5.0) across multiple assessment cycles, yielding a composite score of 0. Full methodology.
DRC / Democratic Republic of Congo(2026-07-31) — July natural Bundibugyo Ebola peak window; 1,003 cases / 254 deaths as of Jun 22 (second-largest outbreak on record);…
EU AI Act — AI Labs Index(2026-08-02) — EU AI Act full applicability for GPAI model obligations, Article 50 transparency, and whistleblower protections (Augu…
Nigeria(2026-08-31) — August 2026 IPC lean-season update — primary checkpoint for Borno State Phase 5 famine trajectory; 34.8M food-insecur…
We reassess nightly.
Special Briefings
Thematic deep-dives: cross-index analysis, structural patterns, and interpretive findings.
Copy-ready citation string for journalism, research, or academic use.
Compassion Benchmark. "Daily Briefing — Jun 23." compassionbenchmark.com/updates/2026-06-23. Accessed [Month Year]. Independent — entities never pay for inclusion, score changes, or suppression of findings.
Daily briefings surface the headline finding. Full benchmark reports include all 40 subdimension scores, complete evidence trails, certified assessments, and sector-level analysis packages — the record researchers and journalists cite.
Independence note: entities never pay for inclusion, score changes, or suppression of findings. Commercial services support access, interpretation, and institutional use only.