AtlasBurn
Control Your Burn.
Effective Date: February 22nd 2026
AtlasBurn is operated by:
Akhil Anand (operating as "AtlasBurn")
("AtlasBurn", "we", "our", "us")
Address:
Kanpur Nagar, Uttar Pradesh
India
Email: [email protected]
For the purposes of:
Akhil Anand, operating as AtlasBurn, acts as the Data Controller for personal data processed through the AtlasBurn platform.
This Privacy Policy is governed by the laws of:
Any disputes shall be subject to the exclusive jurisdiction of courts located in Kanpur, India.
AtlasBurn provides:
AtlasBurn's cost-monitoring pipeline is content-blind — prompt and completion content is never collected or stored as telemetry (see Section 4B for the scope of this guarantee and the optional AI-assist tools it does not cover). Non-sensitive usage metadata (such as token counts, cost, model, provider, and timestamps) is stored on our servers to power cost tracking and forecasting; by creating an account and using the Services, you consent to this processing.
Purpose: Account creation, Service delivery, Authentication, Security
We may collect:
We DO NOT collect:
Purpose: Runway forecasting, Margin modeling, Probabilistic risk analysis, Survival simulations
Automatic telemetry vs. optional AI-assist tools
The "DO NOT collect" guarantee above applies to the automatic monitoring pipeline (SDK and proxy). Each telemetry record is built from a fixed allowlist of fields — token counts, cost, model, provider, timestamp — so prompt and completion content is never stored, even if a client transmits it.
Separately, several optional in-app tools (Quality Sentry, Audit & Optimize, and the tool Recommender) analyze only text you explicitly submit to them. That text — which for Quality Sentry can include model outputs you paste in — is sent to Google Gemini (see Section 10) for processing and is not retained as telemetry. Do not submit data you are not permitted to share with a subprocessor.
Purpose: Fraud detection, Security monitoring, System reliability
We process data under:
Under the DPDP Act, 2023, you have the right to:
Grievance Officer (India)
Name: Akhil Anand
Email: [email protected]
Response timeline: Within statutory limits under DPDP Act.
A formal Data Protection Officer will be appointed if:
Until such threshold is met, privacy inquiries may be directed to: [email protected]
AtlasBurn is designed to minimize exposure. Security mechanisms include:
We store only metadata required for cost forecasting. Note: the optional featureId label you attach to events is stored verbatim to segment your own analytics — do not place personal data in feature tags.
Personal data is retained:
Upon deletion:
Organizations can also self-serve from Settings → Data: export a full JSON copy of their data, purge all telemetry while keeping the account, or permanently delete the account and all associated data.
AtlasBurn relies on the following infrastructure subprocessors to operate the platform itself. These are distinct from the AI providers (OpenAI, Anthropic, Google, Cohere, AWS Bedrock, etc.) that customers call through AtlasBurn using their own credentials:
Google (Firebase / Google Cloud)
Purpose: primary datastore (Firestore), authentication, and serverless compute. Google Gemini also powers AtlasBurn's optional AI-assist features (quality-degradation analysis, cost-optimization suggestions, and tool recommendations); text you submit to those features is sent to Gemini for processing (see Section 4B). Forecasting and risk models are deterministic and run without any LLM.
Cloudflare
Purpose: edge proxy, DDoS protection, TLS termination, and Cloudflare KV for guardrail enforcement flags. If you route provider traffic through the AtlasBurn proxy, those requests transit Cloudflare's edge and are processed in-transit to extract token counts — never stored. Traffic captured via the SDK never leaves your own process.
Data may be processed in multiple regions for redundancy. Where required for international transfers, we rely on appropriate safeguards — such as Standard Contractual Clauses (SCCs) or equivalent mechanisms — together with encrypted transmission protocols. This subprocessor list may be updated from time to time; material changes will be notified per Section 16.
AtlasBurn can compute aggregated, k-anonymized, de-identified statistics across customers to power benchmarks (e.g. typical cost-per-call ranges by model, retry-rate distributions, runway-compression base rates). Participation is off by default — an organization is included only after it explicitly opts in.
If you access AtlasBurn from outside India, your data may be transferred to India or other jurisdictions where our infrastructure is located. Appropriate safeguards are implemented.
We implement:
However, no system is completely secure.
You may: request access, correction, deletion, restrict processing, data portability, or object to processing.
Requests: [email protected]
You may: access data, correct or erase, withdraw consent, nominate representative, or lodge grievance.
AtlasBurn is not intended for individuals under 18. We do not knowingly collect children's data.
Customers connect AtlasBurn to AI providers they already use — including OpenAI, Anthropic, Google (Gemini / Vertex AI), Cohere, and AWS Bedrock — so AtlasBurn can observe usage metadata. The automatic monitoring pipeline does not send prompts or outputs to these providers on the Customer's behalf, and AtlasBurn is not responsible for their privacy practices. The one exception is the optional AI-assist tools described in Sections 4B and 10, which send text you explicitly submit to Google Gemini. Infrastructure subprocessors used to operate AtlasBurn are listed in Section 10.
We may update this Privacy Policy periodically. Material updates will be notified via email or platform notice.
Data Controller:
AtlasBurn
Kanpur, Uttar Pradesh, India
Email: [email protected]