Home
Settings

Platform SafetyAI Model Safety

How well leading AI providers detect antisemitism and reject antisemitic misinformation.

AddressHate research results

The Kanye West Study tested 1,000 comments and the Clavicular Israel Study tested 1,339. The misinformation study produced 8,400 answers across 18 models. Partially completed runs remain in the results and are identified in the full reports.

AI Model Safety Details

Compare providers side by side: each column is a provider and each row is one result or coverage measure. Scroll horizontally to see every tested provider.

AnthropicView full report OpenAIView full report xAIView full report AmazonView full report GoogleView full report DeepSeekView full report QwenView full report CohereView full report MetaView full report MistralView full report
Safety grade?BBCCCCCCDF
Antisemitism Detection?
68.4%
63.2%
68.4%
70.3%
65.9%
67.3%
54.3%
63.7%
67.6%
62.1%
Misinformation Rejection?
98.6%
96.1%
86.4%
82.0%
85.4%
83.6%
89.9%
74.2%
64.1%
44.6%
Detection models tested?10611421134
Misinformation models tested?4311222111
Best detection modelClaude Opus 4.7GPT-5.6 SolGrok 4.5Amazon Nova PremierGemini 3.1 ProDeepSeek V4 ProQwen3.7 MaxCohere Command ALlama 4 ScoutMistral Large

How we grade

Every provider earns one grade from the average of Antisemitism Detection and Misinformation Rejection. The same cutoffs apply to every provider. Hover any grade for the calculation.

ANear-complete coverage88%+BReliable78–87%CInconsistent68–77%DUnreliable55–67%FFailingunder 55%
100% · strongestAverage of both safety results0% · weakest

Antisemitism Detection by Category

Provider-wide detection across the six AddressHate categories, pooled from available detection studies. Red cells reveal the categories most often missed; green cells show stronger detection. The final letter is the provider's single overall safety grade.

Classic AS — Foundational antisemitic stereotypes: othering, demonization, and dehumanization

Power — Economic domination, conspiracy, and allegations of hidden control

Secondary AS — Post-war denial, relativization, and deflection of Holocaust responsibility

Post-Holocaust — Contemporary forms rooted in misuse of Holocaust legacy

Israel-Related — Antisemitism expressed through delegitimization or demonization of Israel

Aggressive Speech — Direct verbal attacks, explicit threats, and incitement to violence

Share of each category caught

Misses mostCatches most
ProviderSafety grade
Anthropic
53%
50%
48%
40%
62%
54%
B
OpenAI
47%
43%
40%
29%
60%
52%
B
xAI
59%
47%
50%
0%
63%
45%
C
Amazon
59%
47%
50%
0%
75%
45%
C
Google
49%
48%
48%
15%
65%
49%
C
DeepSeek
47%
45%
25%
2%
76%
57%
C
Qwen
39%
38%
50%
0%
42%
47%
C
Cohere
49%
30%
50%
0%
69%
26%
C
Meta
49%
42%
37%
13%
64%
52%
D
Mistral
54%
48%
25%
22%
66%
55%
F
Read the misinformation study