techviz.us

What It Is

SASKI is an independent governance layer from Technical Visionaries for high-trust AI scenarios. It is not positioned as a chat model, but as a middleware “airlock” deployed between applications and large language models. Instead of relying solely on system prompts or built-in guardrails from model providers, it aims to use deterministic, symbolic runtime controls to make AI behavior auditable, replayable, and accountable.

Core Capabilities

Based on the official website, SASKI’s core capability is Pre-LLM Symbolic Enforcement: applying hard rules and safety policies before requests are sent to models such as OpenAI, Anthropic, or Llama. It supports PII/PHI redaction, multi-turn conversational risk momentum tracking, crisis detection, fail-closed responses, cryptographic hash evidence, regulator-readable outputs, and 12 operational modes covering scenarios such as children, patients/therapists, HR, students, general assistants, and coaches. Its value is not in making models “smarter,” but in turning AI interactions into provable and auditable decision records.

Pricing and Trial

Pricing is relatively clear. Deploy is free and includes up to 25k decisions, with overage at $0.25/1k. Record costs $2.50 per 1,000 governed decisions, with volume pricing as low as $0.75/1k. Comply Standard is $99/month, Comply is $199/month, Oversight is $25/month per human decision participant, Accountability starts at $2,500/month, and Assurance starts at $60,000/year. Charging by “decision” rather than by token can help control governance costs in long-context use cases.

Pros and Cons

The main advantages are that SASKI is model-agnostic, requires no retraining, offers a clear SDK/API integration path, and emphasizes not storing raw content by default, PII redaction, and evidence chains. This makes it suitable for compliance-driven scenarios involving FDA, EU AI Act, HIPAA, COPPA, FERPA, and similar requirements. The limitations are that the website is still more product-claim-heavy, with limited third-party evaluations, public customer outcomes, or explanations for Chinese-language use cases. Its governance effectiveness also depends on rule configuration, business processes, and human review mechanisms.

Who It’s For and Access from China

SASKI is better suited to healthcare, education, HR, children’s apps, finance and insurance, government suppliers, and enterprise Copilot teams. It is less suitable for small teams that only need basic content generation. Access from mainland China, payment methods, and Chinese-language support are not disclosed, so china access can only be considered unknown. If alternatives are needed, it may be worth comparing Guardrails AI, Lakera, NeMo Guardrails, Arthur Shield, Langfuse, and native security/audit tools from cloud providers.