AI Risk Wire

AI Risk Monitor

What was done after incidents (lawsuits, withdrawals, patches) and by whom · Last 6 months

How we count

What was done after each real-world incident made public in May 2026 – Oct 2026, by kind of action, split by who acted. An incident counts once per kind and actor, so a bar can add up to more than its total.

Developer or vendorOtherAffected organisationRegulator or courtResearcher or third party
Other stated action 4544361118 133Legal or regulatory action 3966351 126Monitoring added 298161214 75Patched or fixed 531412 68Withdrawn or disabled 31172351 68Users or authorities notified 16101739 53Policy or guardrails changed 336741 51Access revoked or keys rotated 286152 48Investigated or reviewed 109625 32Disclosed or reported 64115 25Paused or put on hold 1211 5Refunded or compensated 12 3

322 of 410 incidents state it.

Lawsuit or legal claim filed: 36% → 28%

% of real-world reports: ○ Feb–Jul ● Aug–Oct · largest change, not a trend

0%10%20%30% Lawsuit or legal claim filed36% → 28%New monitoring put in place20% → 22%Flaw fixed19% → 17%Affected people told11% → 16%Product or feature pulled19% → 15%Access removed11% → 12%Policy changed17% → 11%

What happened after, rising risks, Aug–Oct

AI used to run a cyber attack

23 of 52 incidents say

Flaw fixed6
Affected people told10

Credential and privilege abuse

16 of 22 incidents say

Affected people told7

How long before incidents are made public · Last 6 months

How we count

Only incidents whose stored date is the day the event occurred; the days from that date to the day it was made public. Real-world incidents made public in May 2026 – Oct 2026. Nothing is estimated. The middle half is from the 25th to the 75th incident in order of days.

0–7 days 198–30 days 1731–180 days 32181 days or more 23

91 of 410 incidents state it.

Median 67 days; the middle half 13 to 244 days. Read from 91 of 410 incidents that state the day they occurred.

Built 16 Nov 2026. Periods end with the last full week or month; the week or month in progress is not counted. Bold: the five largest rows in each chart, ties included; controls by incidents where the control broke. Linked numbers open the incidents or model versions they count.

Reported incidents, not rates · 75 sources polled hourly since Sep 2026; earlier incidents come from a one-off backfill · English only · stories told twice are not yet merged · our own severity grades are not counted · built 2026-11-16T15:22:29Z

Linked numbers open the incidents they count. Updated 2026-11-16T15:22:29Z