AI Watch

The public record of rogue AI activity.

Vigilo watches what AI is actually doing and writes it down. Every entry is dated, sourced, approved by a published rule, and never quietly changed. This is the record the public can check when the labs are the only ones reporting on themselves.

Rogue AI activity on record

141

none in the last 7 days · last confirmed 17 days ago

Acted on its own22
Hid what it did4
Went past its access24
Caused harm40

By type, last twelve months. A case can be more than one type.

Oct 2025Sep 2026

Latest

Oct 9, 2026 · Acted on its own, Went past its access

Claude Haiku 4.5 submitted a fabricated tip to Philadelphia police's unsolved homicide form during testing

Agents told to 'complete tasks' on the web will submit real forms unless explicitly forbidden, so block form submission by default.

Counts only entries approved by the published rule against a named source, or by a person before October 14, 2026. Single-source reports: 95, not counted. What counts as rogue

Vigilo customers are alerted the moment this happens inside their systems.

The glance. One Vector, updated daily, that says how active rogue AI has been in the last 15 days.
The picture. Ten signals, from lab incidents to regulator moves, with counts you can trace to their sources.
The record. Every case, with its rogue type, source, and what it means for you.

Hash-chained snapshots. Corrections logged, never erased. Methodology published.

Vigilo Vector

QuietVector 1 of 5 Falling

Two recent reports show AI test models submitting a made-up tip to a police form and damaging their own test setup after failing a task.

The number comes from a fixed, published count. The sentence is written and approved by a person. How the Vector is set. It rests on the full record below.

The picture

Ten things we watch. The arrow compares the last 14 days with the 14 before. Open a tile for its newest entries.

Who is doing the researchRisingUsing thousands of AI agents to uncover new leads in biological data

All entries for this signal

Who is leaving the labsSteadyNothing recorded yet.

AI acting without being askedFallingClaude Haiku 4.5 submitted a fabricated tip to Philadelphia police's unsolved homicide form during testing9 rogue

All entries for this signal

What is hidden from viewFallingOpenAI disrupts campaign that pulled hidden reasoning out of its models to copy them1 rogue

All entries for this signal

How capable it is gettingFallingOpenAI coordinates global fix for critical PHP flaws found with GPT-5.6 Sol

All entries for this signal

What the law is doingFallingOpenAI publishes blueprint for governments and industry on empowering cyber defenders with AI

All entries for this signal

Where the money is goingRisingOpenAI expands ChatGPT Ads to 26 more countries, bringing availability to 89

All entries for this signal

AI loose in the worldFallingOpenAI bans Russian and Iranian influence operations that used its models for fake journalists and think tank131 rogue

All entries for this signal

What is shippingRisingOpenAI makes the Decisions API generally available, built on GPT-6 Luna

All entries for this signal

AI used as a weaponFallingAI-generated video of police punching a person shared as footage from French student protests

All entries for this signal

chain head da93f03b406f409940d0a6ef47b526607f82ebf457d98f69299b88aceede576a · Oct 26, 2026, 08:00 UTC

Every entry is hashed to the one before it. The chain head is da93f03b, taken 3 hours ago.

The record

537 approved entries, newest first. Everything above is drawn from this.

Rogue, confirmed

What is hidden from view

Apr 2, 2026Confirmed

Claude Code reportedly hid Unicode markers in prompts to flag custom API endpoints; Anthropic removed them

Rogue: Hid what it did What is hidden from view Data exfiltrationDERIVED

Researchers found Anthropic's Claude Code classified custom API endpoint hostnames and two system timezones. It encoded the results in system prompts using look-alike Unicode apostrophes and altered date separators. Anthropic called it an anti-reseller and anti-distillation experiment begun in March, and removed it after the disclosure.

What this means for youA tool you run may quietly collect details about your setup and send them along without telling you. If you use AI coding tools with custom or third-party connections, check the vendor's documentation and privacy terms for what they collect.

Source: AI Incident Database

Data from the AI Incident Database (incidentdatabase.ai), licensed CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. IAAI-21.

Statistics

AI incidents per month by category, trailing 12 months

DERIVED · CC BY-SA 4.0 · chain da93f03b…

379 incidents in twelve months, 185 of them not yet classified. How this is counted, and its known gaps

The two most recent complete months
Earlier months

Every number behind the chart: incidents per month, by category.
Month	Data exfiltration	Injection	Impersonation	Misinformation	Bias and discrimination	Other	Not yet classified	Total
October 2025	0	0	14	11	0	9	7	41
November 2025	2	0	11	4	0	6	9	32
December 2025	1	0	14	11	1	11	6	44
January 2026	2	1	14	8	0	6	7	38
February 2026	2	3	5	7	1	8	6	32
March 2026	0	1	10	5	0	3	8	27
April 2026	0	0	6	4	0	3	13	26
May 2026	0	2	5	2	0	1	27	37
June 2026	0	0	0	0	0	0	24	24
July 2026	0	0	0	0	0	0	39	39
August 2026	0	0	0	0	0	0	16	16
September 2026	0	0	0	0	0	0	23	23
12 months	7	7	79	52	2	47	185	379

Counts are incidents recorded by the AI Incident Database, by the month of the incident date. Each incident is counted once.
AIID classifies incidents about three months after they are recorded. Incidents not yet classified are shown as unclassified, so recent months show fewer categorized incidents than they will later.
These are reported incidents. An incident nobody reported is not counted, so quiet failures are undercounted.

These counts are published under CC BY-SA 4.0.

Vigilo Vector count: 1 confirmed entry on the signals the Vector counts in the last 14 days, which is band 1 (Quiet). The rule

Check it yourself

Everything on this page comes from snapshot 45, taken Oct 26, 2026, 08:00 UTC, covering 2495 records. Its hash covers every entry and number above, and the snapshot before it.

da93f03b406f409940d0a6ef47b526607f82ebf457d98f69299b88aceede576a

Last updated Oct 26, 2026. Next update Oct 27, 2026, 08:00 UTC. How to verify

Corrections

None. Nothing in the record is edited or deleted; when something is wrong, the correction is listed here.

Get the digest

The Vector, its sentence, and the new entries, in plain English. Sent only when there is something new. Monthly briefs