AI risk
AI attacks and incidents: the record in charts
What has gone wrong with AI, who caused it, whether it was meant, and how that compares with the risks that frameworks describe. Every figure is a count of records in the AI Incident Database or the MIT AI Risk Repository. "Attack" here means an incident coded as deliberate. This page has no list of records: those are on the pages it links to.
Incidents classified under MIT domain 2 (Privacy & Security) or domain 4 (Malicious Actors & Misuse): privacy compromise, attacks on AI systems, fraud and scams, disinformation and surveillance, cyberattacks and weapons.
In brief
678 of the 1,496 classified incidents in the AI Incident Database (45%) fall under the two MIT risk domains for security and malicious use. 243 are dated in the 12 months to February 2026, up 21% on the 201 of the 12 months before. 128 newer incidents are recorded but not yet classified, so comparisons stop at February 2026.
- Incidents in view
- 678
- 45% of the 1,496 classified
- 12 months to February 2026
- 243
- +21% on the 12 months before (201)
- Coded deliberate
- 87%
- of 678 incidents with an intent code
- Not yet classified
- 128
- incidents dated after February 2026
1. Rising: up 21% in the 12 months to February 2026
243 incidents under the security and malicious-use domains are dated in the 12 months to February 2026, against 201 in the 12 months before. The comparison stops at February 2026 on purpose: 128 newer incidents are on record without a classification, and counting those months would show a fall that is only a backlog.
- Security and malicious use
- All recorded incidents
- Classification incomplete
| Month | Security and malicious use | All recorded incidents | Classification |
|---|---|---|---|
| 2023-10 | 11 | 18 | complete |
| 2023-11 | 2 | 10 | complete |
| 2023-12 | 13 | 22 | complete |
| 2024-01 | 8 | 17 | complete |
| 2024-02 | 15 | 27 | complete |
| 2024-03 | 15 | 20 | complete |
| 2024-04 | 15 | 36 | complete |
| 2024-05 | 11 | 25 | complete |
| 2024-06 | 10 | 26 | complete |
| 2024-07 | 15 | 25 | complete |
| 2024-08 | 15 | 24 | complete |
| 2024-09 | 20 | 28 | complete |
| 2024-10 | 18 | 34 | complete |
| 2024-11 | 11 | 18 | complete |
| 2024-12 | 19 | 26 | complete |
| 2025-01 | 20 | 32 | complete |
| 2025-02 | 32 | 47 | complete |
| 2025-03 | 29 | 44 | complete |
| 2025-04 | 29 | 47 | complete |
| 2025-05 | 21 | 34 | complete |
| 2025-06 | 26 | 44 | complete |
| 2025-07 | 21 | 39 | complete |
| 2025-08 | 19 | 35 | complete |
| 2025-09 | 9 | 17 | complete |
| 2025-10 | 19 | 41 | complete |
| 2025-11 | 16 | 28 | complete |
| 2025-12 | 20 | 44 | complete |
| 2026-01 | 20 | 38 | complete |
| 2026-02 | 14 | 30 | complete |
| 2026-03 | 12 | 27 | incomplete |
| 2026-04 | 7 | 25 | incomplete |
| 2026-05 | 7 | 35 | incomplete |
| 2026-06 | 0 | 23 | incomplete |
| 2026-07 | 0 | 33 | incomplete |
| 2026-08 | 0 | 13 | incomplete |
| 2026-09 | 0 | 15 | incomplete |
2. Who was responsible, was it meant, and what kind of harm
The widest path is a person acting with deliberate effect: 398 of the 678 classified incidents in view (59%). Follow a colour from left to right to see where deliberate and accidental incidents end up.
- Deliberate (coded Intentional)
- Accidental (coded Unintentional)
- Unclear (coded Other)
| From | To | Incidents |
|---|---|---|
| AI | Accidental | 63 |
| AI | Deliberate | 192 |
| AI | Unclear | 3 |
| Human | Deliberate | 398 |
| Human | Accidental | 18 |
| Human | Unclear | 1 |
| Other | Deliberate | 1 |
| Other | Unclear | 2 |
| Accidental | Privacy & Security | 55 |
| Accidental | Malicious actors | 26 |
| Deliberate | Malicious actors | 538 |
| Deliberate | Privacy & Security | 53 |
| Unclear | Malicious actors | 2 |
| Unclear | Privacy & Security | 4 |
From accident to attack, year by year
In 2020, 76% of the classified incidents in view were coded deliberate; in 2025, 89%. Deliberate incidents sit at the base of each column, so their growth reads against the whole.
- Deliberate
- Accidental
- Unclear
- Figure over each column: share coded deliberate. * part year.
| Year | Deliberate | Accidental | Unclear |
|---|---|---|---|
| 2019 | 9 | 2 | 0 |
| 2020 | 16 | 5 | 0 |
| 2021 | 9 | 3 | 0 |
| 2022 | 27 | 3 | 0 |
| 2023 | 69 | 7 | 1 |
| 2024 | 159 | 11 | 2 |
| 2025 | 231 | 29 | 1 |
| 2026 | 50 | 8 | 2 |
3. Where incidents concentrate, year by year
The eight kinds of incident with the most records in view. A darker cell is a larger count. The last column is a part year, and its newest incidents are not classified yet.
| Subdomain \ Year | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|---|
| 4.3 Fraud, scams, and targeted manipulation | 3 | 9 | 4 | 13 | 43 | 103 | 190 | 40 |
| 4.1 Disinformation, surveillance, and influence at scale | 2 | 6 | 3 | 6 | 21 | 50 | 38 | 8 |
| 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information | 5 | 5 | 5 | 8 | 9 | 14 | 19 | 4 |
| 2.2 AI system security vulnerabilities and attacks | 1 | — | — | 2 | 3 | 3 | 5 | 7 |
| 4.2 Cyberattacks, weapon development or use, and mass harm | — | 1 | — | 1 | 1 | 2 | 9 | — |
4. What is described against what is recorded
The MIT AI Risk Repository codes 468 risk entries from 74 frameworks to the same subdomains as the incidents in view. Where the two shares differ, either the frameworks are ahead of the record or the record is ahead of the frameworks.
- 2 Privacy & Security
- 4 Malicious actors
- Bubble size: incidents recorded. Dashed line: equal shares. Square-root scale. Unlabelled bubbles are named when pointed at.
| Subdomain | Share of incidents (%) | Incidents | Share of risk entries (%) | Risk entries |
|---|---|---|---|---|
| 4.3 Fraud, scams, and targeted manipulation | 60.9 | 412 | 16.5 | 77 |
| 4.1 Disinformation, surveillance, and influence at scale | 20.4 | 138 | 17.9 | 84 |
| 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information | 13 | 88 | 17.1 | 80 |
| 2.2 AI system security vulnerabilities and attacks | 3.5 | 24 | 23.9 | 112 |
| 4.2 Cyberattacks, weapon development or use, and mass harm | 2.2 | 15 | 17.5 | 82 |
- Incidents recorded
- Risk entries described
Responsible
- A person
- 62%Incidents recorded
- 67%Risk entries described
- An AI system
- 38%Incidents recorded
- 18%Risk entries described
- Other or unclear
- 0%Incidents recorded
- 15%Risk entries described
Intent
- Deliberate
- 87%Incidents recorded
- 69%Risk entries described
- Accidental
- 12%Incidents recorded
- 14%Risk entries described
- Other or unclear
- 1%Incidents recorded
- 17%Risk entries described
Timing
- After release
- 98%Incidents recorded
- 76%Risk entries described
- Before release
- 2%Incidents recorded
- 8%Risk entries described
- Other or unclear
- 0%Incidents recorded
- 16%Risk entries described
5. Who is harmed, and by what
The parties named most often as harmed, or nearly harmed, in the classified incidents in view, and the kind of risk behind each. Read a row as that party's risk profile and a column as the parties one kind of risk reaches. The record names values (epistemic integrity, privacy) as well as people, and a name is an allegation in a report, not a finding.
| Harmed party | 1Discrimination & Toxicity | 2Privacy & Security | 3Misinformation | 4Malicious actors | 5Human-Computer Interaction | 6Socioeconomic & Environmental | 7AI system safety, failures, & limitations | Incidents |
|---|---|---|---|---|---|---|---|---|
| Epistemic Integrity | · | 5 | · | 178 | · | · | · | 183 |
| General Public | · | 10 | · | 151 | · | · | · | 161 |
| Privacy | · | 68 | · | 35 | · | · | · | 103 |
| Democratic Integrity | · | · | · | 62 | · | · | · | 62 |
| National Security And Intelligence Stakeholders | · | 2 | · | 44 | · | · | · | 46 |
| Students | · | 3 | · | 41 | · | · | · | 44 |
| Minors | · | 9 | · | 31 | · | · | · | 40 |
| Women and girls | · | 2 | · | 36 | · | · | · | 38 |
- 1 Discrimination & Toxicity
- 2 Privacy & Security
- 3 Misinformation
- 4 Malicious actors
- 5 Human-Computer Interaction
- 6 Socioeconomic & Environmental
- 7 AI system safety, failures, & limitations
6. Where to look first
The questions a risk owner brings to this record, answered from the counts above. Each answer describes the record, not your exposure: use it to choose which subdomain to open.
Which kind of incident is recorded most?
4.3 Fraud, scams, and targeted manipulation: 163 of the 243 incidents under the security and malicious-use domains dated in the 12 months to February 2026 (67%).
Which is growing fastest?
4.2 Cyberattacks, weapon development or use, and mass harm: 9 incidents in the 12 months to February 2026, against 2 in the 12 months before. The counts are small, so read it as a signal to watch and not as a rate.
What is recorded more often than the frameworks describe?
4.3 Fraud, scams, and targeted manipulation is 61% of the classified incidents here and 17% of the 468 matching risk entries in the MIT AI Risk Repository.
What is described far more than it is recorded?
2.2 AI system security vulnerabilities and attacks: 24% of the risk entries and 3.5% of the incidents (24). A low count is not a low risk: a harm that has not happened yet, or is not reported, leaves no record.
Is deliberate misuse or accident the larger share?
In the 12 months to February 2026, 86% of the classified incidents were coded deliberate (209 of 243), against 93% in the 12 months before.
The board: one card for each kind of incident
Every kind with at least 10 records in view, busiest first. The label on each card is worked out from two counts, by the rules under the cards.
4.3 · Malicious actors
SteadyFraud, scams, and targeted manipulation
- Year to Feb 2026
- 163
- +19% (from 137)
- All years
- 412
- 61% of view
- Deliberate
- 96%
- 17% of risk entries
52 instruments on record for this risk domain
4.1 · Malicious actors
SteadyDisinformation, surveillance, and influence at scale
- Year to Feb 2026
- 42
- -5% (from 44)
- All years
- 138
- 20% of view
- Deliberate
- 96%
- 18% of risk entries
52 instruments on record for this risk domain
2.1 · Privacy & Security
RisingCompromise of privacy by obtaining, leaking or correctly inferring sensitive information
- Year to Feb 2026
- 20
- +33% (from 15)
- All years
- 88
- 13% of view
- Deliberate
- 51%
- 17% of risk entries
16 instruments on record for this risk domain
2.2 · Privacy & Security
EmergingAI system security vulnerabilities and attacks
- Year to Feb 2026
- 9
- +200% (from 3)
- All years
- 24
- 3.5% of view
- Deliberate
- 33%
- 24% of risk entries
16 instruments on record for this risk domain
4.2 · Malicious actors
EmergingCyberattacks, weapon development or use, and mass harm
- Year to Feb 2026
- 9
- +350% (from 2)
- All years
- 15
- 2.2% of view
- Deliberate
- 60%
- 18% of risk entries
52 instruments on record for this risk domain
- Rising
- at least 10 incidents in the last 12 months and up 25% or more on the 12 before.
- Steady
- at least 10 incidents in the last 12 months and within 25% of the 12 before.
- Easing
- at least 10 incidents in the 12 months before and down 25% or more since.
- Emerging
- fewer than 10 incidents in the last 12 months, at least three, and more than in the 12 before.
- Sparse
- too few recent incidents to state a direction.
7. What the rules on record cover
The same seven risk domains twice: the incidents in view, and the policy instruments in this tracker whose recorded use cases fall in that domain. An instrument can count under more than one domain. A long bar on the left beside a short one on the right is a question to ask, not a finding that the rules are missing.
- 1 Discrimination & Toxicity
- 0
- 2 Privacy & Security
- 112
- 3 Misinformation
- 0
- 4 Malicious actors
- 566
- 5 Human-Computer Interaction
- 0
- 6 Socioeconomic & Environmental
- 0
- 7 AI system safety, failures, & limitations
- 0
- 1 Discrimination & Toxicity
- 25
- 2 Privacy & Security
- 16
- 3 Misinformation
- 50
- 4 Malicious actors
- 52
- 5 Human-Computer Interaction
- 56
- 6 Socioeconomic & Environmental
- 142
- 7 AI system safety, failures, & limitations
- 52
How to read this page
- These are reports, not a census. An incident is on record because someone wrote about it and an editor added it. Coverage leans toward English-language media, and a count is not a rate.
- Classification trails the record. The codes for intent, entity and risk domain are added after an incident is recorded. A month is compared only when 75% of its incidents carry a code, which is why the year-on-year figures here end before the newest incident and differ from the calendar figure on AI risk.
- "Deliberate" is a code, not a verdict. It is the MIT causal taxonomy's "Intentional": the outcome was the expected result of pursuing a goal. It covers fraud and impersonation as well as attacks on AI systems themselves.
- Risk entries are counted the same way as the incidents in view. For "Deliberate misuse" only entries coded Intentional are counted; for "Security and malicious use" only entries in domains 2 and 4.
- Updates. Incidents are synced from the AI Incident Database once a day and from its weekly export. Nothing on this page is fetched from the source while you read it.
Disclaimer: informational only, not legal advice. Verify every claim against the linked official sources and consult a qualified lawyer before acting.