AI agents are finding critical CVEs — design your SaaS assuming the server is lost
haltakov · x · 2026-11-11
haltakov argues that AI agents are already finding critical CVEs in software we all run, and 2shell exploits will only multiply. If you run a SaaS, you're responsible for user data even when an attacker gets a shell on your server — so design for that from day one.
His concrete checklist:
- Keep user data encrypted in the DB so a dump yields only ciphertext
- Verify an HMAC on each row to catch unauthorized edits
- Use envelope encryption with key-encryption keys in an HSM
- Enforce a strict policy for releasing data keys
- Decrypt user data only inside a trusted execution environment (TEE)
- Run trusted workloads from a distroless Docker image with SSH disabled; before releasing a data key, verify the workload's cryptographic attestation against the expected image digest
- Deploy the frontend statically from a separate pipeline
Core thesis: treat the largest part of your infrastructure as untrusted, because it can be hacked at any time.
More from coding & agent
- Anthropic's real bet is MCP connectors embedding HubSpot, Adobe and Atlassian inside Claude — petrusenko_max · 2026-11-11
- ZenML Pro launches always-on AI engineering team to keep agents in production — strickvl · 2026-11-11
- Latent Space tests NaiveAI's 2,100 tok/s coding model: fast, verbose, promising — Latent Space · 2026-11-11
- Anaconda's LLMOps guide: retrieval quality degrades independently of the model — anacondainc · 2026-11-11
- Dev demos agent with 'durable object brain' plus on-demand computer, local or remote — samgoodwin89 · 2026-11-11
- Right-sizing intelligence per role: frontier reasoning at the control plane, adversarial verifiers at the gate — clmt · 2026-11-11