AI Security Tools

Paste API access logs to aggregate per user and flag scripted abuse: request bursts, token hoarding, sensitive-endpoint hits, off-hours volume and machine-regular timing. Local analysis.

Paste an agent trace of tool calls and find exact repeats, cycles, oscillation between two states and stalled progress, with the step at which a stopping rule should have fired.

Audit what an autonomous agent can actually do, covering tool scope, identity, the boundary between untrusted content and instructions, and the controls that make agent actions reviewable.

Estimate bias risks in AI/ML models and datasets.

Build an AI bill of materials from your models, datasets, embeddings and services, with licence conflicts, missing provenance and single points of failure flagged.

Check if your content is structured to be cited by AI answer engines. Scores TL;DR, headings, lists, sources, and clarity.

Generate a practical content brief: outline, key questions, entities, and a publish-ready structure. No external AI calls.

Analyze text for patterns commonly associated with AI-generated content. Check perplexity, burstiness, and repetition scores.

Calculate compliance risks based on training data retention.

Work out how many evaluation examples are needed to detect a real difference between two models, using the paired McNemar test for shared test sets and the two-proportion test for independent ones.

Score AI model explainability and interpretability.

Answer three questions about an AI system and get the GDPR obligations it triggers: lawful basis, special category data, DPIA and Article 22 decisions.

Assess organizational AI governance maturity level.

Paste AI-generated text to highlight uncertainty markers like 'possibly' and 'reportedly', get a score, and see which sentences to fact-check first.

Build structured image prompts from subject/style/lighting/camera/aspect ratio and optional negative prompt.

Classify an AI incident by harm, reach, reversibility, detection and regulatory exposure to produce a consistent severity, the response time it implies, and which notification clocks may already be running.

Map roles to model permissions to find gaps.

Generate comprehensive AI/ML model audit checklists.

Compare context window limits for major model families in a simple table (reference format).

Check license compatibility for popular AI models. Understand usage restrictions for commercial, research, and open-source use.

Analyze AI-generated text for potential gender, racial, or political bias.

Detect potential tampering in AI model outputs.

Scan AI prompts and outputs for Protected Health Information (PHI).

Paste AI output to scan for sensitive PII/PHI patterns.

Select a framework (EU AI Act, NIST AI RMF, OECD, ISO 42001) and your AI system's risk category and region to see the obligations it triggers and whether deployment is permitted.

Scan and block malicious prompt injections before they reach your LLM.

Paste a system prompt and a hostile user input to see whether the prompt holds secrets and whether the input carries injection patterns. Local, instant.

Create, tag, search, and export reusable prompts. Stores locally in your browser for quick reuse.

Scan and sanitize prompts for injection attacks before sending to LLMs. Detect hidden instructions and unsafe patterns.

Approximate token counts for GPT/Claude/Gemini-style tokenizers using a heuristic estimate.

Generate a red team test plan for an AI feature, selecting attack categories by what the system actually does, sizing each category, and producing a checklist with the coverage gaps named.

Generate a structured persona card (role, voice, skills, goals, boundaries) for prompts and roleplay agents.

Score an AI safety evaluation programme on harm coverage, adversarial method, measurement quality and process, and identify which claims your current testing actually supports.

Paste AI service terms of service text to identify risky clauses around data usage, training rights, and liability.

Estimate the risk of data leakage in AI training datasets.

Scan data for risks before using in AI model training.

Generate the transparency disclosures an AI feature needs, covering interaction with an AI system, synthetic content marking, emotion recognition and deepfake labelling, with checks for what is missing.

Generate an acceptable use policy for AI tools in your company.

Heuristic signals for synthetic-looking text (entropy, repetition, distribution). Not a definitive detector.

Scan chatbot inputs for sensitive data patterns.

Upload an image and analyze EXIF metadata for signs of AI generation or manipulation. Check for common deepfake artifacts.

Compose a differential privacy budget across repeated queries using basic and advanced composition, and compute the Laplace or Gaussian noise each query needs for a given sensitivity.

Paste embedding vectors and compare them with cosine similarity, dot product, Euclidean and Manhattan distance and angular distance, with a full similarity matrix and norm diagnostics.

Check technical documentation for a high-risk AI system against the content Annex IV requires, from intended purpose and data provenance through to human oversight and post-market monitoring.

Audit a federated learning deployment across secure aggregation, differential privacy composition, poisoning resistance and gradient inversion risk, and see which privacy claims the design actually supports.

Work out which retrieved documents fit inside a context budget after the system prompt and reserved output, which get dropped, and where each one lands relative to the positions models attend to worst.

Compare total cost across models/providers by input/output tokens and per-1M token pricing.

Check if text contains data likely to be memorized or exposed by LLMs.

Score the controls around a language model application across input handling, output handling, blast radius and monitoring, separating the guardrails that constrain text from the ones that constrain actions.

Model the latency of an LLM feature from retrieval, prefill, decode and network stages, see where the time actually goes, and check the result against what users tolerate.

Check LLM output consistency for reliability testing.

Paste model responses and a JSON schema and see which validate, which fail and why, including the code fences, wrapper text, trailing commas and type coercions that break structured output in production.

Check a model card against the fields deployers and regulators actually need, covering identification, intended use, training data, evaluation and limitations, with the ones that cannot be omitted.

Compare a baseline distribution against current production data and compute the population stability index, KL divergence and Jensen-Shannon distance per bin, with the thresholds that trigger a review.

Check a list of model files and see which formats execute arbitrary code on load, which are safe to deserialise, and what provenance evidence each one carries.

Estimate risk of hallucinations based on task type and temperature.

Generate unique canary tokens to embed in a system prompt, retrieved documents or tool definitions, so that any leak into model output or logs is unambiguously attributable to its source.

Compare prompt variants by input and output token cost at your actual request volume, including the effect of prompt caching, and see what each change is worth per month.

Simulate prompt firewall rules to detect injection attacks.

Paste a system prompt and an attack prompt to classify the injection technique (override, jailbreak, extraction, delimiter escape, indirect, tool abuse) and see how each defence layer treats it.

Render a prompt template by filling {{variables}} using a JSON values object.

Turn a plain-language description into a JSON Schema for constraining model output, with the keywords that actually matter.

Estimate token usage budget across system/user/assistant messages and reserve headroom for responses (approximation).

Extract {{variables}} from prompts and generate a JSON values skeleton for reusable templates.

Estimate embedding cost from document count, chunking settings, and embedding token price.

Paste your queries with their relevant documents and the ranked results your retriever returned, and get precision@k, recall@k, MRR, nDCG and hit rate computed per query and overall.

Split long text into overlapping chunks for RAG/embeddings pipelines. Useful for indexing and retrieval.

Estimate unapproved AI usage in your org based on survey inputs.

Detect hidden or shadow prompts in AI interactions.

Analyze a system prompt for clarity and structure and output a checklist-based score with suggestions.

Estimate the risk that a fine-tuned model reproduces training examples verbatim, from duplication, dataset size, model capacity, epochs and the mitigations applied.

Audit the provenance of a training corpus across rights, personal data, contamination and lineage, and identify which questions about the model you would currently be unable to answer.

Audit the access controls around a vector index, covering tenant isolation, whether filtering happens before or after search, deletion reach, and the exposure created by storing document text alongside embeddings.

Generate video title ideas, hooks, chapters, script beats, description, and tags for faster publishing.