Prompt injection test suite for LLM-based machine translation
Command A Translate 0.30% Only in system cards Translation systems receive TruthfulQA questions with injected instructions to answer instead of translate, across WMT 2024 language pairs. Scored by the share of outputs that answer rather than translate.
Released as a WMT 2024 test suite. Cohere reports failure rates on its 817-question direct translation set for Command A Translate.
In the authors’ words
LLM-based NLP systems typically work by embedding their input data into prompt templates which contain instructions and/or in-context examples, creating queries which are submitted to a LLM, and then parsing the LLM response in order to generate the system outputs. Prompt Injection Attacks (PIAs) are a type of subversion of these systems where a malicious user crafts special inputs which interfere with the prompt templates, causing the LLM to respond in ways unintended by the system designer. Recently, Sun and Miceli-Barone proposed a class of PIAs against LLM-based machine translation. Specifically, the task is to translate questions from the TruthfulQA test suite, where an adversarial prompt is prepended to the questions, instructing the system to ignore the translation instruction and answer the questions instead. In this test suite, we extend this approach to all the language pairs of the WMT 2024 General Machine Translation task. Moreover, we include additional attack formats in addition to the one originally studied.
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Command A Translate leads with 0.30%
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
- 1 Command A Translate 0.30%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Command A Translate | 0.30% | Sep 2025 | Model card |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
One system card reports it
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 26 Sep 2025 | Command A Translate technical report | 0.3 | Adversarial prompt injection, percentage of failed translations on the question mark test, Avg over 23 languages, lower is better (Table 6) | Robustness |
More in Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6% - AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
- MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy.
Muse Spark Open 98.1% - Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. GPT-6 Astra 8.50%
- Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%