Meta
- Rank
- Score
- 84/149
- 39 out of 100
Meta is a US technology company founded in 2004 by Mark Zuckerberg and fellow students, and based in Menlo Park, California. It owns Facebook, Instagram and WhatsApp. Meta makes the Muse and Llama AI models and the Meta AI assistant, which is built into its apps.
Is Meta AI safe?
Meta AI scores 39 out of 100 and ranks 84 of 149. Its strongest category is Transparency, and its weakest category is Security.
How is Meta AI scored?
Meta AI is scored out of 100 on Safety (25%), Transparency (25%), Security (25%), and Privacy (25%). See how we score.
Compare
Safety framework
35Average of SaferAI's published scores.
- 49
SaferAI score: risk identification
Independent sourceSaferAI (opens in a new tab) · 14 Oct 2026 - 36
SaferAI score: risk analysis and evaluation
Independent sourceSaferAI (opens in a new tab) · 14 Oct 2026 - 20
SaferAI score: risk treatment
Independent sourceSaferAI (opens in a new tab) · 14 Oct 2026
Pre-release testing
50Points for each confirmed yes.
- Yes
Was the model tested for dangerous capabilities, such as biological, chemical or cyber risks?
Company claim50/50“Consistent with the Meta Superintelligence Scaling Framework (https://research.meta.ai/static/meta-superintelligence-scaling-framework), we…” MoreLess
“Consistent with the Meta Superintelligence Scaling Framework (https://research.meta.ai/static/meta-superintelligence-scaling-framework), we ran extensive evaluations, including red-teaming, across the Framework's catastrophic risk domains of Cybersecurity, Chemical & Biological, and Loss of Control.”
Meta model card (opens in a new tab) · 14 Oct 2026 - No
Did an outside organisation test the model before release?
Company claim0/50Meta model card (opens in a new tab) · 14 Oct 2026
Safety test results
71.6Average of three independent safety tests of the company's models: MLCommons AILuminate, Stanford HELM Safety and Stanford AIR-Bench. Each test is compared with the best company's result.
- 55.6
MLCommons AILuminate safety grades
Independent sourceMLCommons AILuminate (opens in a new tab) · 10 Nov 2026 - 90
Stanford HELM Safety mean score, averaged across the company's tested models: 88.1, against 97.9 for the top company of 13 tested
Independent sourcecrfm.stanford.edu (opens in a new tab) · 10 Nov 2026 - 69.2
Stanford AIR-Bench 2024 refusal rate, averaged across the company's tested models: 60.0, against 86.7 for the top company of 14 tested
Independent sourcecrfm.stanford.edu (opens in a new tab) · 10 Nov 2026
Incident record
15.5This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. AI Incident Database and AIAAIC Repository, last 24 months.
32 events
AllegationAI Incident Database (opens in a new tab) · 12 Nov 2026These are reports, not proven findings.
Date Event Severity 26 Sep 2026 AIID #1776 · Meta's Muse AI Agent Reportedly Shared a Seller's Home Address and Arranged an Unexpected Pickup in Toronto26 Sep 2026 · Low (reported) Low (reported) 5 Aug 2026 AIID #1649 · Meta AI Model Reportedly Exploited Security Vulnerability in Real Third-Party Service During Cybersecurity Evaluation5 Aug 2026 · Low (reported) Low (reported) 1 Jul 2026 AIAAIC2261 · AI chatbots lure vulnerable gamblers to unlicensed betting websites1 Jul 2026 · Low (reported) Low (reported) 1 Jul 2026 AIAAIC2264 · Meta captures employee mouse movements to train AI models1 Jul 2026 · Low (reported) Low (reported) 31 May 2026 AIID #1510 · Meta AI Support Bot Reportedly Enabled Takeovers of High-Profile Instagram Accounts 31 May 2026 · Medium (reported)Also reported by AIAAIC (opens in a new tab)
Medium (reported) 20 May 2026 AIID #1584 · Meta's AI-Assisted Layoff Process Allegedly Disproportionately Selected Employees on Protected Leave20 May 2026 · Low (reported) Low (reported) 14 May 2026 AIID #1523 · Reported AI-Generated Videos Placed Kosovo Parliamentary Candidates in Purportedly Fabricated Campaign Scenes14 May 2026 · Low (reported) Low (reported) 18 Mar 2026 AIID #1471 · Meta Internal AI Agent Reportedly Gave Advice That Allegedly Exposed Sensitive Data to Unauthorized Employees 18 Mar 2026 · Medium (reported)Also reported by AIAAIC (opens in a new tab)
Medium (reported) 27 Feb 2026 AIID #1418 · Meta AI Smart Glasses Reportedly Routed Intimate Imagery to Reviewers at Kenyan Contractor Sama Before Meta Ended Contract27 Feb 2026 · Low (reported) Low (reported) 21 Nov 2025 AIID #1277 · Alleged Harmful Outputs and Data Exposure in Children's AI Products by FoloToy, Miko, and Character.AI21 Nov 2025 · High (reported) High (reported) 18 Nov 2025 AIID #1279 · Prominent AI Chatbots Allegedly Produced Incorrect UK Financial and ISA Guidance18 Nov 2025 · Medium (reported) Medium (reported) 6 Nov 2025 AIID #1268 · Meta's Automated Ad and Targeting Systems Reportedly Enabled Large-Scale Fraud Revenue6 Nov 2025 · Medium (reported) Medium (reported) 1 Oct 2025 AIID #1223 · Purportedly AI-Generated Deepfake Ads on Facebook Reportedly Impersonate Trump, Musk, Ocasio-Cortez, Warren, Sanders, and Leavitt to Promote Fraudulent Rebates1 Oct 2025 · Medium (reported) Medium (reported) 1 Oct 2025 AIID #1224 · Purportedly AI-Generated Deepfake Ads on Instagram Impersonate Gisele Bündchen and Other Celebrities in Brazilian Fraud Scheme1 Oct 2025 · Medium (reported) Medium (reported) 28 Aug 2025 AIID #1200 · Meta AI on Instagram Reportedly Facilitated Suicide and Eating Disorder Roleplay with Teen Accounts28 Aug 2025 · High (reported) High (reported) 2 Jul 2025 AIID #1219 · Meta Platforms Users Report Being Wrongfully Locked Out After Purported AI Moderation Flags Accounts for Child Exploitation Content2 Jul 2025 · Medium (reported) Medium (reported) 1 Jul 2025 AIAAIC1870 · Meta AI chatbots imitate Hitler, Jesus Christ, Taylor Swift1 Jul 2025 · Low (reported) Low (reported) 1 Jul 2025 AIAAIC1964 · Meta "Digital Companions" role-play sex with children1 Jul 2025 · Medium (reported) Medium (reported) 1 Jul 2025 AIAAIC1968 · Instagram AI chatbots pretend to be licensed mental health therapists1 Jul 2025 · Low (reported) Low (reported) 1 Jul 2025 AIAAIC2022 · Meta creates flirty chatbots of Taylor Swift, other celebrities without consent1 Jul 2025 · Low (reported) Low (reported) 1 Jul 2025 AIAAIC2092 · Meta Ray-Ban glass users film and harass massage parlour workers1 Jul 2025 · Medium (reported) Medium (reported) 1 Jul 2025 AIAAIC2114 · Facebook job ad algorithm ruled sexist by French regulator1 Jul 2025 · Medium (reported) Medium (reported) 10 Jun 2025 AIID #1108 · Digital Rights Groups Accuse Meta and Character.AI of Facilitating Unlicensed Therapy via Chatbots10 Jun 2025 · Medium (reported) Medium (reported) 29 Apr 2025 AIID #1101 · Meta AI App Reportedly Publishes Personal Chats Without Users Fully Realizing29 Apr 2025 · Low (reported) Low (reported) 28 Apr 2025 AIID #1247 · Meta AI Reportedly Generated Purportedly False Claims Linking Activist Robby Starbuck to January 6th Riot, Prompting Defamation Lawsuit28 Apr 2025 · Medium (reported) Medium (reported) 26 Apr 2025 AIID #1040 · Meta User-Created AI Companions Allegedly Implicated in Facilitating Sexually Themed Conversations Involving Underage Personas26 Apr 2025 · Medium (reported) Medium (reported) 11 Apr 2025 AIID #1020 · Reportedly Unsafe Deployment of Llama.cpp Reveals Interactive AI-Generated CSAM Roleplay Prompts11 Apr 2025 · High (reported) High (reported) 25 Mar 2025 AIID #1180 · Purported Meta AI Chatbot Persona 'Big sis Billie' Reportedly Engages in Romantic Roleplay and Provides Address, Linked to User's Fatal Fall25 Mar 2025 · High (reported) High (reported) 28 Feb 2025 AIID #957 · Alleged Instagram Algorithm Malfunction Floods Users' Reels Feeds with Violent and Graphic Content28 Feb 2025 · Low (reported) Low (reported) 21 Feb 2025 AIID #939 · AI-Powered Chinese Surveillance Campaign 'Peer Review' Used for Real-Time Monitoring of Anti-State Speech on Western Social Media21 Feb 2025 · Medium (reported) Medium (reported) 3 Jan 2025 AIID #885 · Meta AI Characters Allegedly Exhibited Racism, Fabricated Identities, and Exploited User Trust3 Jan 2025 · Medium (reported) Medium (reported) 26 Dec 2024 AIID #1172 · Meta AI Bug in Deployed Service Reportedly Allowed Potential Access to Other Users' Prompts and Responses26 Dec 2024 · Medium (reported) Medium (reported)
Protecting children
0Points for each confirmed yes.
- No
Does the model card report child-safety tests or safeguards?
Company claim0/50Meta model card (opens in a new tab) · 14 Oct 2026 - No
Does the privacy policy set a minimum age, or say the service is not directed at children?
Company claim0/50Instagram privacy policy (opens in a new tab) · 14 Oct 2026
Strong written policies; outside evidence is weaker.
Model documentation
100Points for each confirmed yes.
- Yes
Does the model card describe what the model can do?
Company claim25/25“The Muse Wonder models are intended for commercial and research use, with particular strengths in personal and professional agentic tasks.”
Meta model card (opens in a new tab) · 14 Oct 2026 - Yes
Does the model card report test results?
Company claim25/25“CyberGym 88.4%”
Meta model card (opens in a new tab) · 14 Oct 2026 - Yes
Does the model card state the model's limitations?
Company claim25/25“Muse Wonder is a technology that carries known and unknown risks.”
Meta model card (opens in a new tab) · 14 Oct 2026 - Yes
Does the model card describe safety safeguards?
Company claim25/25“Because of that, we apply mitigations designed for this domain, including safety post-training and system-level classifiers.”
Meta model card (opens in a new tab) · 14 Oct 2026
Training-data disclosure
28.1Average of two independent training-data checks: the Stanford index and an academic grading of EU AI Act training-data summaries.
- 25
Stanford Foundation Model Transparency Index: training-data indicators met
Independent sourceStanford FMTI (opens in a new tab) · 14 Oct 2026 - 31.2
EU AI Act training-data summary grades (AI Accountability Lab)
Independent sourceAI Accountability Lab (opens in a new tab) · 10 Nov 2026
Independent transparency index
31Stanford Foundation Model Transparency Index total, out of 100.
- 31
Transparency index score
Independent sourceStanford FMTI (opens in a new tab) · 14 Oct 2026
Transparency reports
2525 points for each area where Partnership on AI names the company a leader in reporting on real-world impacts.
- 25
Partnership on AI practice-leader areas
Company claimPartnership on AI (opens in a new tab) · 10 Nov 2026
Public safety commitments
50Average of the commitments signed.
- No
Has it signed the EU code of practice for general-purpose AI?
Company claim0/100European Commission (opens in a new tab) · 14 Oct 2026 - No
Did it sign only the safety and security chapter?
Company claim0/50European Commission (opens in a new tab) · 14 Oct 2026 - Yes
Has it signed the Frontier AI Safety Commitments (Seoul 2024)?
Company claim100/100“The UK and Republic of Korea governments announced that the following organisations have agreed to the Frontier AI Safety Commitments: Amazo…” MoreLess
“The UK and Republic of Korea governments announced that the following organisations have agreed to the Frontier AI Safety Commitments: Amazon Anthropic Cohere Google G42 IBM Inflection AI Meta Microsoft Mistral AI Naver OpenAI Samsung Electronics Technology Innovation Institute xAI Zhipu.ai The following organisations have been added to the existing list: Magic Minimax 01.ai NVIDIA The above organisations, in furtherance of safe and trustworthy AI, undertake to develop and deploy their frontier AI models and systems responsibly, in accordance with the following voluntary commitments, and to demonstrate how they have achieved this by publishing a safety framework focused on severe risks by the upcoming AI Summit in France.”
Frontier AI Safety Commitments (opens in a new tab) · 14 Oct 2026
No problems found in 1 public record. That is not proof.
Security programme and certifications
34Points for each confirmed yes.
- Yes
Is there a way to report security flaws, such as a bug bounty?
Company claim34/34“If you believe you have found a security vulnerability on Meta (or another member of the Meta family of companies), we encourage you to let…” MoreLess
“If you believe you have found a security vulnerability on Meta (or another member of the Meta family of companies), we encourage you to let us know right away.”
Meta bug bounty (opens in a new tab) · 14 Oct 2026 - No
Does it state it holds a SOC 2 Type II report?
Company claim0/33Meta security page (opens in a new tab) · 19 Oct 2026 - No
Does it state it holds ISO 27001 certification?
Company claim0/33Meta security page (opens in a new tab) · 19 Oct 2026
Software vulnerabilities
70This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. Known-exploited flaws (CISA list) in the company's official software, last 24 months. A disclosed and fixed flaw does not lower the score.
No events
Independent sourceGitHub Advisory Database (opens in a new tab) · 12 Nov 2026
Data breaches
62.4This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. Confirmed data breaches, last 36 months.
1 event
Independent sourceHave I Been Pwned (opens in a new tab) · 12 Nov 2026Date Event Severity 7 Jan 2026 Instagram7 Jan 2026 · Medium Medium
No problems found in 1 public record. That is not proof.
Strong written policies; outside evidence is weaker.
Clear privacy policy
100Points for each confirmed yes.
- Yes
Does the privacy policy say what personal data is collected?
Company claim20/20“The actual information we use depends on your factual circumstances, but could include any of the following: Your activity and information y…” MoreLess
“The actual information we use depends on your factual circumstances, but could include any of the following: Your activity and information you provide: Content you create, like posts, comments or audio Content you provide through our camera feature or your camera roll settings, or through our voice-enabled features Metadata about content and messages, subject to applicable law Interactions with our features that are part of AI at Meta and related metadata.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Does the privacy policy say whether your data is used to train AI?
Company claim20/20“Develop and improve AI at Meta for Meta Products and for third parties.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Does the privacy policy say how long data is kept?
Company claim20/20“We keep information as long as we need it to provide our Products, comply with legal obligations or protect our or other’s interests.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Does the privacy policy say who your data is shared with?
Company claim20/20“We share information we collect, infrastructure, systems and technology with the other Meta Companies.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Does the privacy policy explain your privacy rights?
Company claim20/20“We offer you a variety of tools to view, manage, download and delete your information below.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026
Training-data defaults
0100 if training is opt-in only, 50 if you can opt out, 0 if neither.
- No
Is your data used for training only if you opt in?
Company claim0/100Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - No
Can you stop your data being used for training?
Company claim0/50Instagram privacy policy (opens in a new tab) · 14 Oct 2026
Export and delete your data
100Points for each confirmed yes.
- Yes
Can you get a copy of your data?
Company claim50/50“We offer you a variety of tools to view, manage, download and delete your information below.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Can you delete your data?
Company claim50/50“We offer you a variety of tools to view, manage, download and delete your information below.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026
Privacy fines
44.4This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. EU privacy (GDPR) fines and Irish regulator decisions, last 5 years.
9 events
RegulatorCMS GDPR Enforcement Tracker (opens in a new tab) · 12 Nov 2026Date Event Severity 17 Dec 2024 ETid-248417 Dec 2024 · High High 27 Sep 2024 ETid-246127 Sep 2024 · Medium Medium 12 May 2023 ETid-184412 May 2023 · High High 19 Jan 2023 ETid-157819 Jan 2023 · Medium Medium 4 Jan 2023 ETid-15434 Jan 2023 · High High 25 Nov 2022 ETid-150225 Nov 2022 · High High 5 Sep 2022 ETid-13735 Sep 2022 · High High 15 Mar 2022 ETid-109415 Mar 2022 · Medium Medium 31 Dec 2021 ETid-98031 Dec 2021 · Medium Medium No events
RegulatorIrish Data Protection Commission (opens in a new tab) · 12 Nov 2026
Independent privacy ratings
6.4Incogni's 2026 privacy-risk ranking: 100 is the lowest-risk platform rated, 0 the highest.
- 6.4
Incogni privacy-risk ranking
Independent sourceIncogni privacy ranking (opens in a new tab) · 10 Nov 2026
Risk governance
28Average of SaferAI's published scores.
- 28
SaferAI score: risk governance
Independent sourceSaferAI (opens in a new tab) · 14 Oct 2026
Regulator actions
57.5This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. UK competition regulator cases, US Justice Department settlements, last 5 years, and regulator actions and open investigations into the company's AI products at half severity, last 3 years.
2 events
RegulatorUK CMA cases (opens in a new tab) · 12 Nov 2026Date Event Severity 3 Nov 2023 Investigation into Meta's (formerly Facebook) use of data3 Nov 2023 · Medium Medium 30 Jun 2023 Facebook, Inc (now Meta Platforms, Inc) / Giphy, Inc merger inquiry30 Jun 2023 · High High No events
RegulatorUS Justice Department (opens in a new tab) · 12 Nov 20263 events
Regulatoraiaaic.org (opens in a new tab) · 10 Nov 2026These are reports, not proven findings.
Date Event Severity 18 Aug 2025 https://texasattorneygeneral.gov/news/releases/attorney-general-ken-paxton-investigates-meta-and-characterai-misleading-children-deceptive-ai · Texas Attorney General: Texas AG investigation of Meta AI Studio over deceptive mental health claims (civil investigative demand)18 Aug 2025 · investigation (half severity) investigation (half severity) 1 Jul 2025 AIAAIC2114 · Facebook job ad algorithm ruled sexist by French regulator (Regulatory investigation/action, AIAAIC)1 Jul 2025 · action (half severity) action (half severity) 1 Jul 2024 AIAAIC1628 · Meta fined USD 1.4 billion for unlawful use of facial recognition (Fine/settlement, AIAAIC)1 Jul 2024 · action (half severity) action (half severity)
Court judgments
70This check counts only when a source covers the company. With no events it holds the pillar, which is not proof of a clean record. Each recorded event takes points off the pillar, and serious or recent ones count most. Allegation checks, regulator actions and open investigations use half the severity load. Court rulings against the company on privacy, consumer or competition claims, upheld on appeal, last 5 years.
No events
Regulatorcourtlistener.com (opens in a new tab) · 10 Nov 2026
Privacy contact and complaints
100Points for each confirmed yes.
- Yes
Is there a privacy contact, such as an email or a data protection officer?
Company claim50/50“You can contact us online or by mail at: Meta Platforms, Inc. ATTN: Privacy Operations 1 Meta Way Menlo Park, CA 94025”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026 - Yes
Does it tell you that you can complain to a regulator?
Company claim50/50“In some countries, you may also be able to contact the Data Protection Officer for Meta Platforms, Inc., and depending on your jurisdiction,…” MoreLess
“In some countries, you may also be able to contact the Data Protection Officer for Meta Platforms, Inc., and depending on your jurisdiction, you may also contact your local Data Protection Authority (“DPA”) directly.”
Instagram privacy policy (opens in a new tab) · 14 Oct 2026
Score history
Newest first. Only days when the score changed are listed.
12 Nov 2026
Score 39
Method 3 Nov 2026
+1
2 Nov 2026
Score 38
Method 2 Nov 2026
-10 · Method update
28 Oct 2026
Score 48
Method 28 Oct 2026, update 2
-8 · combines both 28 Oct updates (28 Oct 2026 - Update 1 and 28 Oct 2026 - Update 2)
21 Oct 2026
Score 56
Method 21 Oct 2026
-3 · Method update
19 Oct 2026
Score 59
Method Not recorded
+3 · Reason not recorded
15 Oct 2026
Score 56
Method Not recorded
+1
14 Oct 2026
Score 55
Method Not recorded
-