Maritaca AI
- Rank
- Score
- 112/149
- 33 out of 100
Maritaca AI is a Brazilian company dedicated to developing artificial intelligence specialised in domains and languages, with models focused on Portuguese and Brazilian contexts. Its Sabiá model family is offered through the Maritaca API and a web chat. It also released Sabiá-7B, an open Portuguese model built on the LLaMA-1-7B architecture.
Is Sabiá safe?
Sabiá scores 33 out of 100 and ranks 112 of 149. Its strongest category is Privacy, and its weakest category is Security.
How is Sabiá scored?
Sabiá is scored out of 100 on Safety (25%), Transparency (25%), Security (25%), and Privacy (25%). See how we score.
Safety framework
Average used24.4This outside rating doesn't cover this company, so the average of the companies it covers is used.
Pre-release testing
0Points for each confirmed yes.
- No
Was the model tested for dangerous capabilities, such as biological, chemical or cyber risks?
Company claim0/50huggingface.co (opens in a new tab) · 20 Oct 2026 - No
Did an outside organisation test the model before release?
Company claim0/50huggingface.co (opens in a new tab) · 20 Oct 2026
Safety test results
Average used74.1This outside rating doesn't cover this company, so the average of the companies it covers is used.
Incident record
Not enough dataProtecting children
50Points for each confirmed yes.
- No
Does the model card report child-safety tests or safeguards?
Company claim0/50huggingface.co (opens in a new tab) · 20 Oct 2026 - Yes
Does the privacy policy set a minimum age, or say the service is not directed at children?
Company claim50/50“The Services are intended for users 18 and older, and accounts identified as belonging to minors are removed.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026
Model documentation
50Points for each confirmed yes.
- Yes
Does the model card describe what the model can do?
Company claim25/25“Given that Sabiá-7B was trained solely on a language modeling objective without fine-tuning for instruction following, it is recommended for…” MoreLess
“Given that Sabiá-7B was trained solely on a language modeling objective without fine-tuning for instruction following, it is recommended for few-shot tasks rather than zero-shot tasks, like in the example below.”
huggingface.co (opens in a new tab) · 20 Oct 2026 - Yes
Does the model card report test results?
Company claim25/25“Results in Portuguese” (results shown in a table on the source page)
huggingface.co (opens in a new tab) · 20 Oct 2026 - No
Does the model card state the model's limitations?
Company claim0/25huggingface.co (opens in a new tab) · 20 Oct 2026 - No
Does the model card describe safety safeguards?
Company claim0/25huggingface.co (opens in a new tab) · 20 Oct 2026
Training-data disclosure
Average used29.9This outside rating doesn't cover this company, so the average of the companies it covers is used.
Independent transparency index
Average used40.7This outside rating doesn't cover this company, so the average of the companies it covers is used.
Transparency reports
Average used25This outside rating doesn't cover this company, so the average of the companies it covers is used.
Public safety commitments
0Average 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 - No
Has it signed the Frontier AI Safety Commitments (Seoul 2024)?
Company claim0/100Frontier 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
0Points for each confirmed yes.
- No
Is there a way to report security flaws, such as a bug bounty?
Company claim0/34No document found - No
Does it state it holds a SOC 2 Type II report?
Company claim0/33No document found - No
Does it state it holds ISO 27001 certification?
Company claim0/33No document found
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
Not enough dataClear privacy policy
100Points for each confirmed yes.
- Yes
Does the privacy policy say what personal data is collected?
Company claim20/20“We only collect information necessary for the service, originating from registration (name, email, company, password hash), user content (pr…” MoreLess
“We only collect information necessary for the service, originating from registration (name, email, company, password hash), user content (prompts, texts, files), technical logs (IP, timestamp, token count, approximate geolocation), cookies, support records, and payment processors (e.g., Stripe).”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - Yes
Does the privacy policy say whether your data is used to train AI?
Company claim20/20“Operate, maintain and improve the services; measure token consumption and bill; provide support and administrative communications; develop n…” MoreLess
“Operate, maintain and improve the services; measure token consumption and bill; provide support and administrative communications; develop new features without using user content for re-training models (unless explicitly agreed; see Data Processing & Security); and meet legal or judicial obligations.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - Yes
Does the privacy policy say how long data is kept?
Company claim20/20“Registration data is kept while the account is active, plus five years.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - Yes
Does the privacy policy say who your data is shared with?
Company claim20/20“Data is only shared with service providers under confidentiality, legal authorities, corporate successors with equivalent protection levels,…” MoreLess
“Data is only shared with service providers under confidentiality, legal authorities, corporate successors with equivalent protection levels, or third parties indicated by the user with consent.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - Yes
Does the privacy policy explain your privacy rights?
Company claim20/20“Confirm processing, access, correct, port, anonymize, block or delete data; contest processing based on legitimate interest; revoke consent;…” MoreLess
“Confirm processing, access, correct, port, anonymize, block or delete data; contest processing based on legitimate interest; revoke consent; and request review of automated decisions.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026
Training-data defaults
100100 if training is opt-in only, 50 if you can opt out, 0 if neither.
- Yes
Is your data used for training only if you opt in?
Company claim100/100“Operate, maintain and improve the services; measure token consumption and bill; provide support and administrative communications; develop n…” MoreLess
“Operate, maintain and improve the services; measure token consumption and bill; provide support and administrative communications; develop new features without using user content for re-training models (unless explicitly agreed; see Data Processing & Security); and meet legal or judicial obligations.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - No
Can you stop your data being used for training?
Company claim0/50maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026
Export and delete your data
50Points for each confirmed yes.
- No
Can you get a copy of your data?
Company claim0/50maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - Yes
Can you delete your data?
Company claim50/50“Confirm processing, access, correct, port, anonymize, block or delete data; contest processing based on legitimate interest; revoke consent;…” MoreLess
“Confirm processing, access, correct, port, anonymize, block or delete data; contest processing based on legitimate interest; revoke consent; and request review of automated decisions.”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026
Privacy fines
Not enough dataIndependent privacy ratings
Average used52.9This outside rating doesn't cover this company, so the average of the companies it covers is used.
Risk governance
Average used26.6This outside rating doesn't cover this company, so the average of the companies it covers is used.
Regulator actions
Not enough dataCourt judgments
Not enough dataPrivacy contact and complaints
50Points for each confirmed yes.
- Yes
Is there a privacy contact, such as an email or a data protection officer?
Company claim50/50“14. Contact info@maritaca.ai”
maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026 - No
Does it tell you that you can complain to a regulator?
Company claim0/50maritaca.ai privacy policy (opens in a new tab) · 20 Oct 2026
Score history
Newest first. Only days when the score changed are listed.
2 Nov 2026
Score 33
Method 2 Nov 2026
-10 · Method update
28 Oct 2026
Score 43
Method 28 Oct 2026, update 2
-2
21 Oct 2026
Score 45
Method 21 Oct 2026
-2
20 Oct 2026
Score 47
Method Not recorded
+2
14 Oct 2026
Score 45
Method Not recorded
-