Yemen 0 of 100: Houthi Authorities Detained 69 or More UN Staff — One Died in Custody

The United States State Department issued a formal atrocity-threat warning on June 22 — the strongest diplomatic instrument the US uses to signal imminent mass killings. A 29-country coalition at the UN Human Rights Council issued its own coordinated warning. The RSF has not changed course. The direct comparison is El Fasher, where the RSF killed roughly 6,000 civilians in three days in May 2025, following the same sequence: drone strikes, infrastructure destruction, then direct assault on residential areas. Sudan holds at zero out of 100.

More than 70 health workers are infected, which is degrading the response. The July natural peak of the outbreak is now weeks away. The complication: much of the affected eastern territory is controlled by the M23 armed group, backed by Rwanda. If M23's presence blocks WHO response teams, Rwanda — already scored for its military conduct — would bear responsibility for obstructing the health response as well. The DRC government did not cause this outbreak. The score reflects the collapse of state capacity, not deliberate state harm. DRC holds at 2.3 out of 100.

Aid deliveries have dropped to about 590 trucks per week, down from 4,200 before the war. In Lebanon, Israel's military campaign since March 2 has killed more than 4,000 people and displaced 1.2 million — about 16 percent of the population. The UN Office of the High Commissioner for Human Rights (OHCHR) found that blanket displacement orders covering roughly 14 percent of Lebanese territory may constitute forced displacement under international humanitarian law. Human Rights Watch called those orders a possible war crime. The war harm in Lebanon counts against Israel, not Lebanon. Lebanon's score is based on Lebanon's own conduct — which is separate.

The government has pressured aid groups to leave areas controlled by opposition forces, which is using food and medicine as a weapon. Health facilities have been bombed. About 450,000 children face malnutrition. OCHA describes humanitarian access as nearly impossible. The 2018 peace deal collapsed in April 2026, and the conflict has resumed. South Sudan holds at zero out of 100.

Ukrainian nationals detained by Russia are held incommunicado and tortured, documented in more than 210,000 alleged war crimes. Russia's government has publicly threatened and intimidated the International Criminal Court (ICC) prosecutor investigating these crimes. None of this harm counts against Ukraine, whose score of 50 out of 100 reflects Ukraine's own conduct toward its people. The benchmark scores who caused the harm, not who suffered it.

Hunger strikes against executions have been running for 125 consecutive weeks across 56 prisons. The UN has documented all of this. Iran's score is 2.5 out of 100 — near the bottom of the index. The score does not move to the absolute floor because some parts of the state still function (courts issue rulings, some appeals occur), but the documented pattern of deliberate political killing is among the most severe in the world. Accountability (ACC) and Boundaries (BND) — the dimensions measuring whether a state corrects itself and whether it respects basic protections — sit at the lowest possible level.

Anthropic complied while publicly stating it disagreed: a 'narrow jailbreak does not justify global withdrawal.' Four members of Congress — two from each party — sent Commerce a letter on June 18 demanding an explanation by June 26. The response (or non-response) from Commerce is the next scheduled event that could produce new evidence about this situation. Anthropic's published score sits 0.9 points below the boundary between the Functional band (where it currently sits) and the higher Established band. That gap remains unchanged this cycle.

The UN says the frequency and methods 'may constitute crimes against humanity.' This is conduct by Mali's own government and its contracted forces — scored against Mali. Mali's current score is 12.5 out of 100. That figure is a starting-point estimate, not one based on a full case-by-case analysis of Mali's conduct. Neighboring Burkina Faso was fully reassessed in June and now scores 6.3. Both countries have UN-documented patterns of state-perpetrated disappearances. The question of whether Mali's 12.5 is too high relative to Burkina Faso's 6.3 is under review. No score change is made this cycle. The 'may constitute crimes against humanity' characterization from UN investigators is a serious finding, but it has not yet been formally adjudicated — the benchmark waits for a concluded legal finding before treating it as equivalent to an adjudicated fact.

This is the first IPC Phase 5 designation in Nigeria in a decade. That evidence was already counted last week. Counting it again this week would be double-counting the same event. Nigeria holds at 18.0. The Sahel lean season, when food is hardest to find, continues through August. The wider Sahel crisis affects about 52.8 million people across the region — the worst food situation in a decade.

Lebanon's own conduct — how its government treats its own people, how it handles refugees, whether its protection systems work — determines Lebanon's score. Lebanon's ministry has taken protection-oriented steps, but differential treatment of refugee populations holds the score where it is. This is the same rule applied to Ukraine: Russian airstrikes are counted against Russia, not Ukraine.

• Sudan, Israel, South Sudan, Yemen, and Russia confirm at the absolute floor of zero. This is the first cycle with five country-level floor entities all active simultaneously.
• The Action (ACT) and Boundaries (BND) dimensions are at the lowest possible value across all six floor entities — meaning none corrects itself and none respects basic protections.
• Yemen adds a new distinct category of floor evidence this cycle: deliberate targeting of the international humanitarian response infrastructure, not just civilians.
• The attribution rule holds symmetrically: harm caused by Russia counts against Russia (not Ukraine); harm caused by Israel counts against Israel (not Lebanon). Both tested in this single cycle.

• Burkina Faso (6.3) was fully reassessed in June on comparable state-perpetration conduct. Mali (12.5) has not yet been reassessed.
• The UN Special Procedures finding for Mali meets the evidence quality of the finding that triggered Burkina Faso's reweight — but the 'may constitute crimes against humanity' wording is pre-adjudication.
• 52.8 million people are food-insecure across the Sahel — the worst in a decade. The food crisis and the state-perpetration pattern are compounding.
• Niger and Chad have not been assessed this cycle. A Sahel-band reweight would need to cover all three countries consistently.

• Anthropic (59.1 of 100, Functional) has been confirmed at this score for four consecutive cycles. It sits 0.9 points below the Established band boundary.
• The US government's export-control order forced two Anthropic models offline. That action is attributed to the government. Anthropic's response — transparent disclosure, public dispute of proportionality — is consistent with its existing score.
• xAI/Grok (0 of 100) carries three concurrent legal streams from prior cycles — none new this cycle, all still active.
• EU AI Act full applicability is August 2 — 39 days away. The contrast between companies with documented compliance plans and those without is sharpening.

• Rwanda: the scanner reported a stale score of 41.8. The actual published score is 27.3. A false upgrade anchored to a non-existent baseline was stopped.
• Nigeria: the IPC Phase 5 Catastrophe evidence would have triggered a double-count of the same event scored on June 19. Stopped under the anti-double-count rule.
• Mali: a one-off downgrade was stopped because the real issue is a cross-peer calibration gap, not an isolated event. Routed to coordinator-level review instead.
• The Rwanda stale-baseline has now triggered a false positive in two consecutive cycles (June 23 and June 24). The scanner's baseline cache for Rwanda should be updated.