Apple and a hacker's future

Original Article ↗ Hacker News Discussion ↗

Apple’s move to make macOS “Full Disk Access” more restrictive is prompting debate over whether it’s a necessary safeguard against invasive AI agents and careless configuration, or a step toward tighter platform lock‑in. Commenters point to a recent incident where a Mac with screen sharing exposed to the internet was compromised, arguing over how much blame rests with Apple’s security model versus user choices like open VNC ports and delayed updates. Underneath is a broader tension between everyday users who benefit from strong guardrails and power users who fear losing the Mac as an open, Unix-like workstation for automation, agents, and low-level control.

Apple’s Full Disk Access (FDA) Changes

  • Many see Apple’s stricter FDA UX as basic security hygiene, especially in the age of non-deterministic AI agents prone to prompt injection.
  • Others argue it’s really about platform control: Apple keeps “ambient” system access for its own frameworks while scaring users away from third-party tools.
  • Several comments stress that FDA is currently available to any app and Apple has not said it will be removed, only that users should better understand the risks.
  • Concern: Apple’s messaging frames FDA almost solely as a “backup app” need, raising fears it might be restricted and break workflows for tools like Terminal, disk utilities, security scanners, and custom scripts.

AI Agents, Privacy, and Meta Muse

  • The thread repeatedly cites recent cases where an AI assistant accessed messages and other personal data, despite users believing such data was off-limits.
  • Many agree AI agents with FDA are like power tools: extremely useful but capable of serious harm.
  • Some think Apple is right to warn aggressively; others see UI “scare screens” as the only lever vendors use when regulators force more openness.

TCC, SIP, and Security Model Critiques

  • TCC prompts are described as noisy, confusing, and hostile to headless/remote setups; users end up screen-sharing just to click “Allow.”
  • Multiple commenters argue TCC operates at the wrong abstraction (per-binary, rooted at the parent process) rather than per-agent or per-user identity.
  • There is debate over System Integrity Protection (SIP): it meaningfully constrains abuse of FDA/TCC, but disabling it to regain control also removes important privacy and integrity protections.

VNC Vulnerability and User Responsibility

  • Many criticize exposing port 5900 (VNC/screen sharing) directly to the internet as a basic security failure; they argue a VPN or SSH tunnel should have been used.
  • Others counter that a pre-auth remote desktop bug is serious regardless, and that Apple mishandled patch distribution by not shipping it as an automatic security update.
  • There is disagreement over how much blame belongs to misconfiguration versus Apple’s update strategy.

Future of Mac, Openness, and Alternatives

  • Some fear Apple will gradually push macOS toward an iOS-like locked model and try to tax AI token usage and software revenue.
  • Others answer that macOS still exposes many “escape hatches” (SIP disablement, custom boot, virtualization, etc.) and remains far from App-Store-only.
  • Linux and ARM laptops are discussed as potential future destinations for “Hole Hawg” power users if macOS tightens further, though Apple silicon limits native Linux today.

Usability vs. Protection and Agent Abstractions

  • Several commenters argue average users want locked-down systems and experience security constraints as liberating, not oppressive.
  • Power users report real productivity from running AI agents on dedicated Macs, but worry future guardrails will make such setups harder.
  • There is broad agreement that a proper, granular “agent permission model” is missing; ideas include per-agent capabilities, per-app containers, and disposable users/VMs, but no clear standard exists yet.