How Syntic keeps conversations safe
Syntic combines automated safeguards with human review. Automated systems look for harmful, abusive or deceptive content; flagged conversations can be reviewed by people; and decisions with serious consequences are never left to an automated flag alone.
Automated safeguards
During sign-up, Syntic tells every new user that automated safeguards protect conversations from harmful, abusive or deceptive content. A model may decline a request that breaks the usage policy.
Safety review
A conversation can be flagged for safety review. According to the Privacy Policy, Syntic can use flagged conversations to get better at detecting harmful content and enforces its policies, even if you have switched off Help improve Syntic. See Model training and your chats.
Human oversight
- Syntic’s terms say an automated flag alone does not lead to a report to law enforcement.
- Limited disclosure may be considered only in exceptional cases involving an imminent threat of death or serious injury, where the law allows, with human review when feasible.
- Restrictions on an account are meant to be proportionate, with notice and a chance to respond where practicable. See Account enforcement and appeals.
Your own safety settings
Settings → Safety has a Trusted contact section. You can invite an adult you trust (they must be 18 or older). The page explains that Syntic may, in the future, notify that person if a conversation suggests a serious safety concern. Notifications are not active yet.
For AI Staff and agents
When Syntic acts for you, the risks are different. Organizations decide what each AI worker may access and do, and should require human approval for consequential actions where the product supports it. The deployment safety page describes the approach and its limits: safeguards reduce risk but can’t remove it.