Security Model

PAI implements defense-in-depth across four layers. The goal is not just privacy but operational security — making it difficult for an adversary to determine what you did, when, or where.

Layer 1: Amnesic by design

Like Tails, PAI is amnesic. It runs entirely from RAM and the read-only squashfs on the USB drive.

• No hard disk access — the host machine’s drives are never mounted
• RAM-only operation — all state exists only in volatile memory
• Clean shutdown — powering off clears all RAM:
  • Chat history and AI conversations
  • Downloaded files and browser history
  • Wi-Fi credentials and network logs
  • Every keystroke and mouse movement

There is no forensic residue. The machine returns to its pre-boot state.

Layer 2: Network anonymity

MAC spoofing randomizes your hardware address on every boot. The access point, ISP, and any local observer sees a different device each time.

The firewall denies all incoming connections. Ollama and PAI UI listen only on localhost — they are not reachable from the network, even on the local subnet.

Layer 3: Local-only AI

The AI runs entirely on-device using Ollama. No tokens, no API keys, no cloud dependencies.

• Model execution: CPU-only (GPU libs stripped to save space)
• Network: Ollama binds to 127.0.0.1:11434 and never phones home
• Supported models: Any GGUF-compatible model — Llama 3, Mistral, Phi, Gemma, Qwen, DeepSeek, and more

Layer 4: Hardened defaults

✓ MAC randomized on every boot
✓ Firewall: deny incoming, localhost only
✓ No SSH server enabled by default
✓ GPG configured with SHA-512 + AES-256
✓ Tor pre-configured (opt-in)
✓ WireGuard available for VPN tunnels
✓ LUKS available for encrypted volumes
✓ No telemetry, no analytics, no phoning home — ever
✓ Firefox ESR with hardened policies
✓ Automatic memory wipe on shutdown

Threat model

PAI protects against:

• Network surveillance (MAC spoofing + Tor + encrypted DNS)
• Forensic analysis of the host machine (amnesic, no disk writes)
• Cloud AI data harvesting (all inference is local)
• Tracking across sessions (new MAC, clean browser, no cookies)
• Local network attacks on AI services (firewall blocks all incoming)

PAI does NOT protect against:

• Physical keyloggers or hardware implants on the host machine
• Compromised BIOS/UEFI firmware
• Shoulder surfing or screen recording
• Adversaries who control your ISP and the Tor exit node simultaneously
• Cold boot attacks on RAM (mitigated by full shutdown, not sleep)
• A compromised USB drive — always verify ISO checksums before flashing

Comparison with cloud AI