Who handles your data
Abdou Kaba operates AfterCast, a platform for evaluating software agents against recorded incidents. You can contact him about privacy or a workspace data request at amkaba@usfca.edu.
Google sign-in and account data
When you choose Google, Google authenticates you and Cloudflare Access checks whether you may enter AfterCast. The workspace receives a verified account identifier and email address. AfterCast uses the identifier to create and reopen your workspace, associate your records with it, and keep users’ records separate. It displays your email in the signed-in interface. The account table stores the identifier and the account creation time.
Google handles your password. The AfterCast application does not request access to Gmail messages, contacts, calendars, or Drive files. It uses Google identity data for authentication and account administration. It does not sell that information, use it for advertising, or send it to model providers as part of sign-in.
What the workspace stores
We store information you submit or generate, including incident evidence, logs, datasets and their revisions, prompts or outputs included in a case, evaluation scores, human reviews, and trace spans. Traces can include identifiers, service names, timestamps, status, and attributes supplied by your instrumentation. This information can contain personal data if you include it in your submissions.
AfterCast uses these records to replay cases, compare agents, investigate runs, and display your results. Cloudflare hosts the service and its D1 workspace database. When you create a trace-ingestion key, the database keeps a digest of the key and its workspace association.
Cookies, hosting, and integrations
These public pages use no advertising or analytics cookies. The private workspace uses Cloudflare Access session cookies to keep you signed in. Cloudflare processes connection and security information, which can include your IP address and request time, to deliver and protect the service.
Enabling an integration changes where the relevant data goes. A live Claude evaluation sends the case question and requested evidence to Anthropic. An incident import retrieves the issue and selected logs from the source you connect. A configured Temporal bridge receives workflow inputs, and a configured Tempo archive receives trace data. Provider keys supplied for a live evaluation or import are used for that request. Avoid including credentials in the evidence itself.
Providers may process data outside your country. Their own services and accounts are subject to their respective terms and privacy notices. See the Google Privacy Policy and the Cloudflare Privacy Policy.
Purpose and retention
We process account and workspace information to provide the service you request. We process limited operational information to keep the service available, investigate errors, and protect accounts. Where data protection law requires a legal basis, we rely on performance of the service agreement for requested features and our legitimate interest in maintaining a reliable, secure service for operational processing.
Saved evaluations and datasets persist between sessions so you can reproduce earlier runs. To request removal of an account and its stored records, email the privacy contact from the address associated with the workspace and specify what you want removed. We may verify ownership before acting and retain information where legally required. Copies you exported and archives operated by your organization have separate retention arrangements.
Your choices and rights
Depending on your location, you may request access, correction, deletion, portability, or restriction of processing. You may also object to certain uses and lodge a complaint with your local data protection authority. If an organization submitted your information in an incident, contact that organization or the privacy contact above so we can identify the relevant workspace. Please send identifiers rather than confidential incident logs with your initial request.