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.
Every incident in the AI Incident Database.
In brief
The AI Incident Database records 1,708 incidents, 1,496 of them classified with the MIT risk taxonomy. 441 are dated in the 12 months to February 2026, up 29% on the 341 of the 12 months before. 128 newer incidents are recorded but not yet classified, so comparisons stop at February 2026.
- Recorded incidents
- 1,708
- in the AI Incident Database
- 12 months to February 2026
- 441
- +29% on the 12 months before (341)
- Coded deliberate
- 49%
- of 1,496 incidents with an intent code
- Not yet classified
- 128
- incidents dated after February 2026
1. Rising: up 29% in the 12 months to February 2026
441 recorded AI incidents are dated in the 12 months to February 2026, against 341 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.
- Classified
- All recorded incidents
- Classification incomplete
| Month | Classified | All recorded incidents | Classification |
|---|---|---|---|
| 2023-10 | 18 | 18 | complete |
| 2023-11 | 9 | 10 | complete |
| 2023-12 | 21 | 22 | complete |
| 2024-01 | 15 | 17 | complete |
| 2024-02 | 26 | 27 | complete |
| 2024-03 | 19 | 20 | complete |
| 2024-04 | 33 | 36 | complete |
| 2024-05 | 25 | 25 | complete |
| 2024-06 | 23 | 26 | complete |
| 2024-07 | 25 | 25 | complete |
| 2024-08 | 22 | 24 | complete |
| 2024-09 | 28 | 28 | complete |
| 2024-10 | 33 | 34 | complete |
| 2024-11 | 17 | 18 | complete |
| 2024-12 | 26 | 26 | complete |
| 2025-01 | 30 | 32 | complete |
| 2025-02 | 45 | 47 | complete |
| 2025-03 | 43 | 44 | complete |
| 2025-04 | 47 | 47 | complete |
| 2025-05 | 34 | 34 | complete |
| 2025-06 | 39 | 44 | complete |
| 2025-07 | 34 | 39 | complete |
| 2025-08 | 30 | 35 | complete |
| 2025-09 | 13 | 17 | complete |
| 2025-10 | 34 | 41 | complete |
| 2025-11 | 23 | 28 | complete |
| 2025-12 | 38 | 44 | complete |
| 2026-01 | 31 | 38 | complete |
| 2026-02 | 26 | 30 | complete |
| 2026-03 | 20 | 27 | incomplete |
| 2026-04 | 13 | 25 | incomplete |
| 2026-05 | 10 | 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 an AI system acting with accidental effect: 681 of the 1,496 classified incidents in view (46%). 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 | 681 |
| AI | Deliberate | 271 |
| AI | Unclear | 8 |
| Human | Deliberate | 463 |
| Human | Accidental | 58 |
| Human | Unclear | 4 |
| Other | Accidental | 1 |
| Other | Unclear | 9 |
| Other | Deliberate | 1 |
| Accidental | AI system safety, failures, & limitations | 288 |
| Accidental | Privacy & Security | 55 |
| Accidental | Misinformation | 136 |
| Accidental | Human-Computer Interaction | 33 |
| Accidental | Discrimination & Toxicity | 197 |
| Accidental | Socioeconomic & Environmental | 5 |
| Accidental | Malicious actors | 26 |
| Deliberate | Malicious actors | 538 |
| Deliberate | Misinformation | 56 |
| Deliberate | Privacy & Security | 53 |
| Deliberate | Discrimination & Toxicity | 44 |
| Deliberate | Human-Computer Interaction | 9 |
| Deliberate | Socioeconomic & Environmental | 17 |
| Deliberate | AI system safety, failures, & limitations | 18 |
| Unclear | Malicious actors | 2 |
| Unclear | AI system safety, failures, & limitations | 7 |
| Unclear | Misinformation | 4 |
| Unclear | Privacy & Security | 4 |
| Unclear | Discrimination & Toxicity | 2 |
| Unclear | Socioeconomic & Environmental | 2 |
From accident to attack, year by year
In 2019, 30% of the classified incidents in view were coded deliberate; in 2025, 65%. 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 | 13 | 30 | 1 |
| 2020 | 23 | 65 | 2 |
| 2021 | 18 | 60 | 1 |
| 2022 | 40 | 65 | 0 |
| 2023 | 89 | 79 | 1 |
| 2024 | 185 | 104 | 3 |
| 2025 | 265 | 140 | 5 |
| 2026 | 54 | 42 | 4 |
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 |
| 7.3 Lack of capability or robustness | 14 | 31 | 36 | 31 | 27 | 32 | 37 | 13 |
| 3.1 False or misleading information | 1 | 5 | 3 | 10 | 35 | 50 | 66 | 18 |
| 4.1 Disinformation, surveillance, and influence at scale | 2 | 6 | 3 | 6 | 21 | 50 | 38 | 8 |
| 1.1 Unfair discrimination and misrepresentation | 8 | 19 | 8 | 10 | 11 | 7 | 3 | — |
| 1.2 Exposure to toxic content | 4 | 4 | 7 | 10 | 12 | 17 | 21 | 4 |
| 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information | 5 | 5 | 5 | 8 | 9 | 14 | 19 | 4 |
| 5.1 Overreliance and unsafe use | 1 | — | 4 | 5 | 5 | 3 | 18 | — |
4. What is described against what is recorded
The MIT AI Risk Repository codes 1,600 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. Across 24 subdomains the two rankings agree moderately (Spearman's rank correlation 0.46).
- 1 Discrimination & Toxicity
- 2 Privacy & Security
- 3 Misinformation
- 4 Malicious actors
- 5 Human-Computer Interaction
- 6 Socioeconomic & Environmental
- 7 AI system safety, failures, & limitations
- 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 | 27.6 | 412 | 4.8 | 77 |
| 7.3 Lack of capability or robustness | 20.4 | 305 | 7.9 | 126 |
| 3.1 False or misleading information | 12.8 | 192 | 3.3 | 53 |
| 4.1 Disinformation, surveillance, and influence at scale | 9.2 | 138 | 5.3 | 84 |
| 1.1 Unfair discrimination and misrepresentation | 7.9 | 118 | 5.2 | 83 |
| 1.2 Exposure to toxic content | 6.1 | 91 | 7.3 | 116 |
| 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information | 5.9 | 88 | 5 | 80 |
| 5.1 Overreliance and unsafe use | 2.5 | 38 | 3.8 | 60 |
| 1.3 Unequal performance across groups | 2.3 | 34 | 1.1 | 17 |
| 2.2 AI system security vulnerabilities and attacks | 1.6 | 24 | 7 | 112 |
| 4.2 Cyberattacks, weapon development or use, and mass harm | 1 | 15 | 5.1 | 82 |
| 6.1 Power centralization and unfair distribution of benefits | 0.4 | 6 | 3.4 | 54 |
| 6.2 Increased inequality and decline in employment quality | 0.4 | 6 | 3.4 | 55 |
| 6.3 Economic and cultural devaluation of human effort | 0.3 | 5 | 1.9 | 31 |
| 7.4 Lack of transparency or interpretability | 0.3 | 5 | 2.6 | 42 |
| 3.2 Pollution of information ecosystem and loss of consensus reality | 0.3 | 4 | 1.4 | 22 |
| 5.2 Loss of human agency and autonomy | 0.3 | 4 | 2.9 | 47 |
| 6.4 Competitive dynamics | 0.2 | 3 | 1.3 | 21 |
| 6.5 Governance failure | 0.2 | 3 | 3.8 | 61 |
| 7.1 AI pursuing its own goals in conflict with human goals or values | 0.2 | 3 | 6.3 | 100 |
| 6.6 Environmental harm | 0.1 | 1 | 3.6 | 57 |
| 7.2 AI possessing dangerous capabilities | 0 | 0 | 4.9 | 78 |
| 7.5 AI welfare and rights | 0 | 0 | 0.2 | 3 |
| 7.6 Multi-agent risks | 0 | 0 | 3.3 | 53 |
- Incidents recorded
- Risk entries described
Responsible
- A person
- 35%Incidents recorded
- 37%Risk entries described
- An AI system
- 64%Incidents recorded
- 42%Risk entries described
- Other or unclear
- 1%Incidents recorded
- 21%Risk entries described
Intent
- Deliberate
- 49%Incidents recorded
- 34%Risk entries described
- Accidental
- 49%Incidents recorded
- 35%Risk entries described
- Other or unclear
- 1%Incidents recorded
- 31%Risk entries described
Timing
- After release
- 98%Incidents recorded
- 61%Risk entries described
- Before release
- 2%Incidents recorded
- 12%Risk entries described
- Other or unclear
- 1%Incidents recorded
- 27%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 |
|---|---|---|---|---|---|---|---|---|
| General Public | 27 | 10 | 61 | 151 | 2 | · | 14 | 265 |
| Epistemic Integrity | 6 | 5 | 63 | 178 | 1 | 3 | 5 | 261 |
| Privacy | 9 | 68 | 1 | 35 | 2 | · | 5 | 120 |
| Students | 11 | 3 | 5 | 41 | 2 | 2 | 22 | 86 |
| Democratic Integrity | 1 | · | 10 | 62 | · | · | · | 73 |
| Women and girls | 27 | 2 | · | 36 | 1 | · | 6 | 72 |
| Minors | 16 | 9 | 2 | 31 | 3 | 1 | 9 | 71 |
| Women | 25 | 2 | · | 26 | 1 | · | 6 | 60 |
- 1 Discrimination & Toxicity
- 2 Privacy & Security
- 3 Misinformation
- 4 Malicious actors
- 5 Human-Computer Interaction
- 6 Socioeconomic & Environmental
- 7 AI system safety, failures, & limitations
- Information and communication
- 81
- Arts
- 35
- Entertainment and recreation
- 35
- Transportation and storage
- 28
- Wholesale and retail trade
- 20
- Law enforcement
- 16
- Human health and social work activities
- 15
- Education
- 15
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 441 recorded AI incidents dated in the 12 months to February 2026 (37%).
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 28% of the classified incidents here and 4.8% of the 1,600 matching risk entries in the MIT AI Risk Repository.
What is described far more than it is recorded?
7.1 AI pursuing its own goals in conflict with human goals or values: 6.3% of the risk entries and 0.2% of the incidents (3). 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, 61% of the classified incidents were coded deliberate (239 of 392), against 66% 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
- 28% of view
- Deliberate
- 96%
- 4.8% of risk entries
52 instruments on record for this risk domain
3.1 · Misinformation
SteadyFalse or misleading information
- Year to Feb 2026
- 64
- +19% (from 54)
- All years
- 192
- 13% of view
- Deliberate
- 28%
- 3.3% of risk entries
50 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
- 9.2% of view
- Deliberate
- 96%
- 5.3% of risk entries
52 instruments on record for this risk domain
7.3 · AI system safety, failures, & limitations
SteadyLack of capability or robustness
- Year to Feb 2026
- 41
- +24% (from 33)
- All years
- 305
- 20% of view
- Deliberate
- 5%
- 7.9% of risk entries
52 instruments on record for this risk domain
1.2 · Discrimination & Toxicity
SteadyExposure to toxic content
- Year to Feb 2026
- 20
- +5% (from 19)
- All years
- 91
- 6.1% of view
- Deliberate
- 20%
- 7.3% of risk entries
25 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
- 5.9% of view
- Deliberate
- 51%
- 5% of risk entries
16 instruments on record for this risk domain
5.1 · Human-Computer Interaction
RisingOverreliance and unsafe use
- Year to Feb 2026
- 15
- +275% (from 4)
- All years
- 38
- 2.5% of view
- Deliberate
- 18%
- 3.8% of risk entries
56 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
- 1.6% of view
- Deliberate
- 33%
- 7% 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
- 1% of view
- Deliberate
- 60%
- 5.1% of risk entries
52 instruments on record for this risk domain
1.1 · Discrimination & Toxicity
SparseUnfair discrimination and misrepresentation
- Year to Feb 2026
- 2
- -71% (from 7)
- All years
- 118
- 7.9% of view
- Deliberate
- 19%
- 5.2% of risk entries
25 instruments on record for this risk domain
1.3 · Discrimination & Toxicity
SparseUnequal performance across groups
- Year to Feb 2026
- 2
- -50% (from 4)
- All years
- 34
- 2.3% of view
- Deliberate
- 12%
- 1.1% of risk entries
25 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
- 243
- 2 Privacy & Security
- 112
- 3 Misinformation
- 196
- 4 Malicious actors
- 566
- 5 Human-Computer Interaction
- 42
- 6 Socioeconomic & Environmental
- 24
- 7 AI system safety, failures, & limitations
- 313
- 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.