AI Safety Assessment
Evaluate your organization's AI systems against the Safer Agentic AI Framework using automated evidence analysis. Your API key connects directly to Anthropic — no data passes through our servers.
Step 1: Connect your API key
Enter your Anthropic API key in the bar above to enable live evaluation. Your key is stored in browser memory only and is never persisted or logged.
Get an API key at console.anthropic.com
What is an API key? Learn more
An API key lets your browser communicate directly with Claude, Anthropic's AI model. It's like a password that authenticates your requests.
To get one: Create a free account at console.anthropic.com, then generate an API key in your account settings.
Cost: Each evaluation costs approximately $0.01-0.03 in API usage. A full sub-goal assessment (3-5 requirements) typically costs under $0.15. Anthropic offers free trial credits for new accounts.
Privacy: Your key stays in your browser's memory only. It is never sent to SaferAgenticAI.org. When you close the tab, the key is gone.
Assessment Workflow
- 1Connect your Anthropic API key using the input bar above. Your key stays in browser memory only.
- 2Browse the 16 safety suites (9 Drivers + 7 Inhibitors) in the sidebar and select the sub-goal you want to evaluate.
- 3Drop files or folders into the evidence area, or paste text directly. PDF, DOCX, TXT, Markdown, and more. All processing happens locally in your browser.
- 4Click "Evaluate" on each evidence requirement. Claude will analyze your evidence and return a score (1–5: Unacceptable / Poor / Average / Good / Excellent), justification, gaps, and relevant excerpts.