Anthropic Already Understands Identity, Retention, and User Control. Now Apply It to Deepfakes.
Anthropic recently updated its privacy policy language.
At first glance, this may sound like a normal privacy update.
But for anyone paying attention to AI deepfakes, AI-generated voices, synthetic images, and digital identity abuse, the update says something much bigger.
It shows that AI companies already understand identity verification.
They already understand data retention.
They already understand third-party data flows.
They already understand user controls.
They already understand that AI systems are moving beyond simple chat and into real-world actions.
So the question is simple:
If AI companies already understand these principles for privacy, safety, and product operations, why are they not applying the same seriousness to AI-generated images, videos, and voices before those outputs are released into the world?
Anthropic says:
“To help keep our services safe and secure, we may ask you to confirm your age or identity.”
Source:
https://privacy.claude.com/en/articles/10301952-updates-to-our-privacy-policy
That matters.
Because one of the objections to source-level accountability is that identity verification is supposedly too extreme, too unrealistic, or too difficult.
But Anthropic’s own privacy-policy update says they may already ask users to confirm age or identity to help keep services safe and secure.
That is the point.
Identity verification is not impossible.
Identity verification is not foreign to AI companies.
Identity verification is already part of the safety conversation.
No Anonymous AI is saying that same principle should apply before unrestricted use, download, posting, export, publication, commercialization, or distribution of realistic AI-generated images, videos, and voices.
Create normally.
Watermark first.
KYC/AML before use.
Original stored.
Accountability attached.
This does not mean stopping creativity.
This does not mean blocking people from using AI.
This means that before realistic synthetic media is released into the world without restrictions, there should be real accountability attached.
Anthropic also says:
“Claude can increasingly carry out longer tasks and work with third-party apps and services.”
Source:
https://privacy.claude.com/en/articles/10301952-updates-to-our-privacy-policy
That matters too.
AI is no longer only a chatbot answering questions.
AI is becoming an action layer.
AI is connecting to apps.
AI is performing longer tasks.
AI is working across services.
AI is moving closer to real-world consequences.
When AI moves from conversation to action, accountability becomes even more important.
If an AI system can interact with third-party services, complete tasks, and move through connected workflows, then the company building that system already knows it must think about responsibility, data flows, permissions, and consequences.
The same logic should apply to synthetic media.
If an AI system can generate a realistic image, video, or voice that can damage a child, a family, a public figure, a company, a school, a bank, or a community, then accountability should not begin after the harm spreads.
It should begin at the source.
Anthropic says its update adds details about:
“what’s shared with a third party when you connect a service, and what we receive in return when Claude completes tasks on your behalf.”
Source:
https://privacy.claude.com/en/articles/10301952-updates-to-our-privacy-policy
Again, that matters.
Anthropic is already describing data flows.
What goes out.
What comes back.
What happens when a service is connected.
What happens when Claude completes a task.
That is a responsibility chain.
No Anonymous AI is asking for a similar responsibility chain for AI-generated images, videos, and voices.
Who generated it?
Was it watermarked?
Was it restricted?
Was it downloaded?
Was it exported?
Was it posted?
Was it published?
Was it commercialized?
Was it released without restriction?
Was the original generated asset stored?
Was accountability attached to the verified person or entity responsible for releasing it?
These are not impossible questions.
They are the same kind of accountability questions AI companies already understand when they describe connected services, third-party data flows, and user controls.
Anthropic also says:
“We don’t sell your data, Claude remains ad-free, and you can control whether your conversations are used to improve Anthropic’s AI models.”
Source:
https://privacy.claude.com/en/articles/10301952-updates-to-our-privacy-policy
That is another important point.
AI companies already know how to build user controls.
They already know how to give users choices around data use.
They already know how to design product settings that affect what happens to user content.
So why not create stronger controls for public release of synthetic media?
If users can control whether conversations are used to improve AI models, then AI companies can also build controls around whether realistic AI-generated images, videos, and voices can be downloaded, exported, posted, published, commercialized, or released without identity accountability.
The issue is not whether controls can exist.
The issue is whether AI companies are willing to create them before more people are harmed.
Anthropic also says:
“If you choose to allow us to use your data to improve Claude, we’ll retain this data for 5 years.”
Source:
https://privacy.claude.com/en/articles/10301952-updates-to-our-privacy-policy
That matters because another objection to No Anonymous AI is that storing original generated assets would be too difficult, too burdensome, or too unusual.
But long-term retention already exists in AI operations.
Anthropic’s own policy update says certain data may be retained for 5 years if users allow it to be used to improve Claude.
So the idea of retaining data is not foreign to AI companies.
No Anonymous AI is asking for original generated assets to be stored as accountability anchors when realistic AI-generated images, videos, or voices are released without restriction.
Not without privacy protections.
Not without legal review.
But as a serious accountability system designed to protect real people from synthetic harm.
Because deepfake abuse is not just “content.”
It is identity abuse.
It is reputation abuse.
It is emotional abuse.
It is financial abuse.
It is public manipulation.
It is sometimes sexual abuse.
It is sometimes a weapon against children.
Removal after harm is not enough.
Detection after harm is not enough.
Apologies after harm are not enough.
The companies building these tools already know how to discuss verification, user control, third-party data flows, and retention.
Now they need to apply those same principles to synthetic media harm.
This is especially important because major AI CEOs already know the broader risks.
Anthropic CEO Dario Amodei said:
“We have to figure out how to make this technology reliably. But the onus primarily falls on us.”
Source:
https://abc7.com/post/anthropic-ceo-dario-amodei-calls-stronger-regulation-ai-exclusive/19275685/
He is right.
The onus primarily falls on the companies building and releasing these systems.
Dario Amodei also wrote:
“AI is advancing at a lightning pace.”
And:
“By contrast, policy—and especially legislation—moves very slowly.”
Source:
https://darioamodei.com/post/policy-on-the-ai-exponential
Again, he is right.
AI is moving faster than government.
But that makes company responsibility more urgent, not less.
If policy is slow, waiting for policy cannot be the only answer.
If legislation moves slowly, AI companies should not use slow legislation as the excuse for slow accountability.
OpenAI CEO Sam Altman warned of a:
“Significant impending fraud crisis.”
He also said:
“A thing that terrifies me is apparently there are still some financial institutions that will accept the voiceprint as authentication.”
And:
“That is a crazy thing to still be doing. AI has fully defeated that.”
Source:
https://www.securityweek.com/openais-sam-altman-warns-of-ai-voice-fraud-crisis-in-banking/
That warning matters.
If AI can defeat voice authentication, then AI-generated voices must be part of the accountability standard.
This is not only about fake images.
This is not only about fake videos.
This is also about fake voices.
Microsoft CEO Satya Nadella said after AI-generated explicit images of Taylor Swift went viral:
“What is our responsibility? Which is all of the guardrails that we need to place around the technology…”
He also said:
“I think we can govern a lot more than we think we give ourselves credit for.”
Source:
https://www.fonearena.com/blog/415134/microsoft-ceo-ai-safeguards-taylor-swift-deepfake-viral.html
That is the right question.
What is the responsibility?
The responsibility is not only to remove abuse later.
The responsibility is not only to write policy papers.
The responsibility is not only to say government should move faster.
The responsibility is to build systems where realistic AI-generated images, videos, and voices cannot be released anonymously without accountability.
No Anonymous AI is not asking for creativity to stop.
We are not saying KYC/AML is the only possible answer.
Our proposed framework may not be perfect. I am certain @elonmusk, harmed public figures, advocates, creators, and others may have even better solutions.
Let’s work together to protect everyone using AI, not just a select group.
But one thing should be clear:
This must be handled at the source.
AI companies can HELP STOP AI deepfake abuse at the source.
They already understand identity.
They already understand retention.
They already understand data flows.
They already understand user control.
They already understand safety and security.
Now apply those principles to AI-generated images, videos, and voices.
Create normally.
Watermark first.
KYC/AML before use.
Original stored.
Accountability attached.
If there is a better framework, please share your ideas and solutions.
Because doing nothing is not a solution.
Waiting for government alone is not leadership.
And protecting people only after harm spreads is not enough.
No Anonymous AI: