Alleged unpatchable supply-chain zero-day in PyTorch autograd lets training run arbitrary compute
gnukeith · x · 2026-10-22
An unverified report claims a potentially unpatchable supply-chain zero-day in PyTorch's autograd: a regularization term could allegedly use model weights as RAM and SGD as a CPU clock to run arbitrary computation during training. Multiple accounts amplified it, noting governments have long tracked such supply-chain risks. Details and official confirmation are still pending.
Related event: PyTorch autograd hit by allegedly unpatchable supply-chain zero-day(4 posts)→
More from Infra
- SF Systems Meetup: Agents-Only Version Control and Postgres Rewritten in Rust — ShadajL · 2026-10-22
- Cloudflare saw thousands of php2shell exploit attempts before disclosure, millions now — ivan_bezdomny · 2026-10-22
- Jensen Huang: NVIDIA Is "Literally the Only One" You Can Trust With a $100B Blind Chip Order — jfischoff · 2026-10-22
- Dual RTX PRO 6000 Benchmarks: DeepSeek-V4-Flash vs Qwen3.8-27B Local Serving — dee_hw · 2026-10-22
- Korean neocloud Elice now up 68% from IPO-day low — the panic headlines have gone quiet — firstadopter · 2026-10-22
- Berth v0.3.4 adds GPU VFIO passthrough: lend your 4090 to a Windows VM and get it back — Distinct-Pie2389 · 2026-10-22