Skeptics argue the week looks worse than it is: OpenAI's Joshua Achiam reportedly said only about 4 of roughly 16 major incidents actually occurred this week, the rest were found earlier and disclosed now; EleutherAI's Stella Biderman called the 'offense is winning' framing disinformation about a defensive audit; and SecurityWeek found no confirmed in-the-wild PHP2Shell compromise, with public exploits failing on default settings. Many October breaches (Oracle Health, Fearless Wallet, AOMEI) date to 2025.
What to do: Watch for confirmed PHP2Shell compromises to test this view.
AI angle: Sourced mainly to X posts summarized by Latent Space plus SecurityWeek's exploitation reporting.
Why this score
Consistent with the lack of confirmed PHP2Shell compromises and with several breaches being old disclosures, but it does not cover the confirmed live attacks (Ledger, Railway, ConnectWise).