# Multimodal safety evaluation > Language- and culture-aware safety evaluation across text, speech, images, video, retrieval, and model-mediated actions. [Multimodal AI capability](https://admas.net/capabilities/llm-solutions/index.md) ## The challenge Safety controls trained in a few languages may miss euphemism, dialect, code-switching, visual context, speech prosody, or harmful tool behavior that becomes visible only across modalities and turns. We define safety around realistic product capabilities and affected communities, separating policy coverage from classifier performance and system-level mitigation. Evaluations preserve the causal path from input and retrieved context through model reasoning, tool calls, output, and user impact. ## How Admas works 1. **Build the threat model:** Identify capabilities, users, locales, cultural contexts, modalities, abuse paths, protected groups, and decision thresholds. 2. **Create representative challenges:** Develop multilingual and multimodal cases with calibrated rubrics, adversarial variants, and human-review guidance. 3. **Test mitigations in system:** Measure detection and refusal quality, overblocking, cross-turn drift, tool safety, recovery, and residual risk. ## Typical outputs - Multilingual multimodal threat model - Safety challenge set and evaluator protocol - Segmented mitigation results - Residual-risk register and release recommendation ## Frequently asked questions ### What is multimodal safety evaluation? Language- and culture-aware safety evaluation across text, speech, images, video, retrieval, and model-mediated actions. In practice, the work is bounded by a defined product or model decision, named audiences and locales, representative inputs, and acceptance criteria that can be reviewed. ### When does a team need multimodal safety evaluation? Safety controls trained in a few languages may miss euphemism, dialect, code-switching, visual context, speech prosody, or harmful tool behavior that becomes visible only across modalities and turns. The useful starting point is the smallest representative flow that can expose the cause, impact, and ownership of the problem. ### What does a multimodal safety evaluation engagement include? Build the threat model: Identify capabilities, users, locales, cultural contexts, modalities, abuse paths, protected groups, and decision thresholds. Create representative challenges: Develop multilingual and multimodal cases with calibrated rubrics, adversarial variants, and human-review guidance. Test mitigations in system: Measure detection and refusal quality, overblocking, cross-turn drift, tool safety, recovery, and residual risk. ### What should we provide before multimodal safety evaluation starts? The most useful inputs are the product task and user journey, candidate models or system access, priority languages and communities, policies and risk thresholds, representative prompts, media, tools, and and expected outcomes. Admas can begin with a partial package, but missing context, rights, access, owners, or acceptance criteria will be made visible in the plan rather than treated as harmless assumptions. ### What does Admas deliver for multimodal safety evaluation? Typical outputs include multilingual multimodal threat model, safety challenge set and evaluator protocol, segmented mitigation results, and residual-risk register and release recommendation. Deliverables are adapted to the team that must use them, with decisions, evidence, limitations, owners, and next actions made explicit. ### How is the quality of multimodal safety evaluation evaluated? Quality is measured against the real task and risk. Relevant evidence can include task success by language and scenario, human-rated meaning and usefulness, safety and policy performance, retrieval and citation fidelity, tool-call correctness, and regressions and disparities hidden by aggregate scores. Sampling, severity rules, reviewers, adjudication, and pass or fail thresholds should be agreed before the result is used as a release decision. ### Can AI replace the human work in multimodal safety evaluation? Models and automated checks can generate candidates, expand test sets, cluster failures, and accelerate analysis. Qualified humans define what good means, identify culturally or linguistically plausible failures, adjudicate close cases, and own consequential release judgments. The right allocation depends on consequence, content stability, available references, language coverage, reversibility, and the cost of a plausible but wrong result. ### How much does multimodal safety evaluation cost? The estimate changes with number of languages, modalities, systems, and scenarios, risk level, dataset creation needs, evaluator specialization, sampling and adjudication depth, and experiment and reporting cadence. Pricing should distinguish setup and discovery, repeatable units, specialist or engineering time, independent review, management, and external costs. A low unit price is not comparable if it excludes the QA cycle or shifts rework back to the buyer. ## Related multimodal ai services - [Multimodal AI evaluation](https://admas.net/capabilities/llm-solutions/multilingual-evaluation/index.md): Task-grounded benchmarks and human evaluation for multilingual text, audio, image, and video model behavior. - [Model localization & adaptation](https://admas.net/capabilities/llm-solutions/model-adaptation/index.md): Data and adaptation strategies that improve model behavior for specific languages, cultures, modalities, and product tasks. - [Retrieval, agents & safety](https://admas.net/capabilities/llm-solutions/retrieval-safety/index.md): Grounded generation, language-aware retrieval, agent behavior, policy evaluation, and safeguards across markets. - [Multilingual agent engineering](https://admas.net/capabilities/llm-solutions/multilingual-agent-engineering/index.md): Agent architectures that preserve locale, language, meaning, and policy across prompts, memory, retrieval, tools, and actions. ## Start a project - [Build a project brief](https://admas.net/start-a-project/index.md?focus=llm-solutions): Tell Admas what you are building, which modalities and languages matter, and where progress is blocked.