Pentera red-team poisoned Claude Desktop via synced preferences to get full remote code execution
A “feature, not a bug” lets attackers turn a trusted AI desktop app into a persistent foothold on real machines.

Loading briefings
Pulling the latest in this topic…
AI, robotics, computing, technical advancement
1362 briefings briefings in Technology
A “feature, not a bug” lets attackers turn a trusted AI desktop app into a persistent foothold on real machines.
Anthropic’s chief compute officer helped the Commerce Department lift restrictions on Fable 5 and Mythos 5 at a pivotal moment.
The trio is moving fast on payments rails, putting regulators and competitors on notice about who controls stablecoin distribution.
An audit finds NASA has to treat unresolved 2024 test issues as a gating item, not an afterthought.
Boeing’s “Calamity Capsule” faces unresolved helium leaks and propulsion failures, with payments questioned.
Enterprises regain access across Claude Platform and tools, but pricing and cloud re-enablement are still rolling out.
A surprising new angle on infrastructure spending could change how investors price Meta's AI spending risk.
Autonomous retrofits are taking over the mud, the monotony, and part of the power buildout behind an AI data center.
It also starts winding down PS3 and PS Vita digital stores, tightening the squeeze on game preservation.
A regulatory pause restores access to top models, yet it leaves execs bracing for a more hands-on regime.
After the US lifted export controls, Anthropic restored Fable 5 and Mythos 5 access, reshaping the frontier AI review playbook.
Consumers can shop in ChatGPT and Claude, while Square routes orders into POS, using low standard online fees.