AI conditions.
How artificial intelligence works in Bephic, what information it uses and what responsibilities you and we have.
Last updated: October 1, 2026
Scope
These conditions supplement the Terms of Service and the Privacy Policy, and apply to all of Bephic's AI features. Bephic is a product of Neuraphic Inc. When we say "we", we mean Neuraphic Inc. and its affiliates or successors.
Bephic uses AI mainly in two ways: the Bephic assistant, which works for you and your team in the panel, and the AI for your customers, which replies on behalf of your business on WhatsApp, Instagram, Messenger, TikTok and Threads.
The AI also writes the text of your posts and drafts replies to the reviews on your Google Business Profile. Automatic review replies are governed by the Terms of Service.
The Bephic assistant
On paid plans, the assistant uses AI models. On the free plan, it answers with fixed rules and does not use AI models.
The assistant can read the files, photos and documents you send it, transcribe what you dictate by voice, write text, create images for your ads, and import your business details from a text, a web page, a PDF or your Google Business Profile.
The assistant proposes each change in a preview. Nothing changes in your account until you, or a person on your team with permission, confirm it. A preview left unanswered changes nothing.
In Permissions, you can authorize Bephic to take certain actions without asking you, such as moving an ad's budget or publishing a draft. All of them ask for confirmation until you authorize them, except appointment bookings, and you can withdraw each authorization at any time.
The AI for your customers
The AI for your customers is available on paid plans. It starts off and replies only on the networks where you turn it on (WhatsApp, Instagram, Messenger, TikTok or Threads). You can turn it off whenever you want.
On TikTok, it only answers direct messages in conversations the person started. On Threads, it posts public replies. Bephic never writes to your customers by email.
The AI replies only with the business data you approved. It can offer times to book, save a prospect's details, note a summary of the conversation and hand it over to your team. It never promises discounts or gifts that are not in that data.
When the conversation passes to a person
- When the customer asks to talk to a person.
- When the customer is upset.
- When a sale exceeds the amount you set.
- When it is missing a fact to answer: it does not make it up.
- When the topic is legal, medical or a problem with a payment.
Notice that it is an automatic assistant
The AI never claims to be a person. On Instagram and Messenger, it says in its first message that it is an automatic assistant. On WhatsApp, it says so in its first message to numbers with the +1 prefix. In all other cases, it says so when the customer asks.
Your customers' information
Your business is responsible for its customers' information. Bephic handles it as a processor, only on your instructions and in line with the Data Processing Agreement. We keep your customers' conversations on the five channels for 30 days and never use them to train AI models.
Bephic is not designed for protected health information under the US HIPAA law, and we do not sign business associate agreements. Bephic does not handle medical information. Your business must not put that information in Bephic, and what your business or its customers send to Bephic is your business's responsibility.
Images and voice
The AI can read the photos that you or your customers send and transcribe voice notes, for the sole purpose of understanding them and replying.
The images the assistant creates are proposals. You decide whether to publish them, and you must review them before doing so.
We review the images uploaded to Bephic and reject content prohibited by the Acceptable Use Policy. We also strip the metadata from every photo uploaded, such as the place where it was taken.
Models and providers
Bephic may use text, image and voice AI models from several providers, for example Amazon Web Services, Google, Microsoft Azure, Anthropic, Meta, OpenAI or Neuraphic Inc.'s own models. We choose providers by their price and quality, and we may change them, or make a Neuraphic Inc. model the main one, at any time and without prior notice.
Each outside provider, that is, anyone other than Neuraphic Inc., receives only the minimum information needed to provide the service and is bound not to train its models on it and not to keep it longer than needed to respond.
Neuraphic Inc. is the exception. It improves its AI models only with what the "Help improve our AI models" option allows, described in Model training, and never with the information of your business's customers. When its models serve Bephic, they receive the same minimum information as any other provider.
The list of providers that process or may process information on Bephic's behalf is in Subprocessors. For AI, your information goes only to companies on that list.
A model created by another company may run inside the infrastructure of one of the companies on that list. In that case, whoever created the model does not receive your information.
We do not sell or rent your information, nor share it for advertising.
No warranty and your responsibility
The AI can make mistakes. It can give an answer that is incorrect, incomplete or out of context. To the extent the law allows, Bephic does not guarantee that what the AI produces is correct or suitable for your business.
Your business is responsible for what the AI says to its customers and for what you publish or approve with it. We ask you to review each preview before confirming it and to keep your business data up to date.
Your business is also responsible for having its customers' permission to write to them and for complying with the laws on commercial messages, such as Law 2300 of 2023 in Colombia.
To the extent the law allows, your business answers to Bephic for claims that arise from its messages, from the way it configures the AI and from its content. The limits of Bephic's liability are in the Terms of Service.
Nothing in these conditions limits the rights that the law of your country gives you.
Model training
There is no switch when you create the account. Before you start, Bephic shows you the "You help us improve" notice. If after reading it you press "Start", you accept the Terms of Service and authorize us to use your information to improve the AI.
We never treat your silence or a pre-ticked box as authorization.
You can turn off this use at any time with the "Help improve our AI models" option, in Settings, Data and privacy. A signed written agreement with Bephic may provide otherwise.
With this option on, the way you use Bephic, including your conversations with the assistant, helps improve the AI of Neuraphic Inc. and its affiliates or successors.
In a workspace, each person has their own option. The owner's option also governs the whole workspace: if the owner has it off, nothing that anyone does inside that workspace is used for training, even if their own option is on.
Whether it is on or not, we never use the following to train AI models:
- Your customers' information, or their messages, photos or voice notes.
- Sensitive information, such as health data.
- The information of accounts whose signed written agreement excludes this use.
- Information received from your Google accounts.
If your conversations with the assistant include your customers' information, we remove it before saving any copy for training.
Before using any information to train AI models, we carry out a dated privacy impact assessment.
If you turn the option off, we stop using your information for training from that moment and delete the copies saved for training within 30 days. What a model has already learned cannot be removed.
How long we keep your conversations with the assistant
With the option on, we keep your conversations with the assistant for 2 years and use them for training. With the option off, we keep them for 30 days, only so the assistant remembers recent context, unless a signed written agreement provides otherwise.
If Bephic is separated from Neuraphic
If Bephic stops being part of Neuraphic Inc., Neuraphic Inc. keeps the models and what they learned. It will receive new information from your account only if you keep this option on.
If the company that operates Bephic changes, we will tell you by email. The details are in the Terms of Service.
The assistant's memory
The assistant remembers facts about your business, such as hours, prices and payment methods, and memories about you and the way you work. You can turn memory off or delete it whenever you want.
- To turn it off, switch off "Use memory" in Settings, Memory. The assistant stops reading and saving memories.
- To delete a fact or a memory, delete it in Settings, Memory.
- To delete everything the assistant remembers about you, use "Reset the assistant" in Settings, Data and privacy.
- To delete all of your business's data, use "Delete what Bephic knows" in Settings, Memory.
The AI for your customers replies with your business data, even if the assistant's memory is off. If you delete that data, the AI will pass your customers' questions to your team until you load it again.
What you delete is removed from Bephic. If it was already used for training while the training option was on, what the model learned stays.
Signed written agreements
If you and Bephic sign a different written agreement, that agreement may provide that we never use your information or your customers' information to train AI models, that there is no option to turn it on, and that your conversations with the assistant are kept for 30 days.
That agreement may also set additional conditions.
Prohibited uses
You may not use Bephic's AI to send spam, commit fraud, harass other people or create or distribute illegal content. Nor may you use it to target children under 13 or to collect their information. The full rules are in the Acceptable Use Policy.
Prohibited content counts as a strike. Spam, scams on customers, harassment, offering prohibited goods and using accounts that are not your business's own also count as a strike, once confirmed after a report. Three strikes in 90 days suspend the account.
Child sexual abuse material blocks the account immediately and we report it to NCMEC, as US law requires.
We close the account directly if a legal or authority order requires it, or if it was created from a country where Bephic is not available. The available countries are listed in Availability.
We also close the accounts of repeat copyright infringers and accounts that belong to minors.
Fraud also blocks the account, for example a card fraud warning, a bank dispute over a charge you did authorize or abuse of connected accounts. A payment reversal that the law grants you, such as the one in Law 1480 of 2011 in Colombia, never blocks the account.
We tell you the cause by email, and you may appeal within 30 days. A person reviews the appeal and answers you within 10 business days at most. If the appeal is accepted, we restore your account and everything it had. The details are in the Terms of Service.
To report misuse, write to [email protected] or follow the steps in Report abuse.
How AI is paid for
On the free plan, the assistant does not use AI models. Each paid plan includes an amount of AI use per month, and what you do not use in the month does not carry over to the next.
When what is included runs out, the AI can keep working with your balance (AI credits), if you allow it. You can top up your balance and turn on automatic top-up, always with your consent.
Only the workspace owner can pay with the card. Admins can request top-ups and lower spending limits, but not raise them.
Plans and balance are always charged in US dollars (USD).
- Balance is not refunded or exchanged for money.
- Balance that you pay for does not expire.
- Balance that Bephic gives, for example in promotions, may expire. The date is shown in the panel.
- All balance is lost if the account is deleted, closed or blocked.
These rules apply unless the law of your country provides otherwise. Refunds are governed by the Terms of Service.
Changes
We may change, suspend or remove AI features.
If we make a material change to these conditions, such as to your rights or to the use of your information, we will tell you in the panel 30 days ahead. Other changes are published on this page with a new update date.
If we raise a price, we will tell you 30 days ahead by email and in the panel, along with how to cancel.
Contact
For questions about your information or model training, write to [email protected]. We reply within 10 business days at most.
To report a vulnerability, write to [email protected]. For any other inquiry, use the support form.
Neuraphic Inc.'s full contact details are in the Privacy Policy.