Tag Archives: Facebook

Attribution is a Myth? Not Any More. (Labeling AI Content on Facebook)

I know that I said that attribution is a myth in Stefan Gladbach’s Christmas video. But I’m trying to fix it on the Bredemarket Facebook page.

Attribution elsewhere

Here on the Bredemarket blog, there are two distinct writers:

  • John E. Bredehoft (human, or so they say), whose posts such as this one are identified as “Bredemarket.”
  • Bredebot (non-human identity powered by Google Gemini skills), whose posts are identified as “Bredebot.”

So it’s fairly simple to keep track of who writes what.

LinkedIn is a little more complicated, but still clear. Anything Bredebot writes is initially posted on the Bredebot showcase page, and when it’s shared elsewhere, it’s tagged at the top with the identifier Bredebot. For anything else originating at a Bredemarket page or showcase page (except for reshared material), assume John wrote it.

Attribution on Facebook

Things kinda sorta parallel on Facebook, except that Facebook groups are the equivalent to LinkedIn showcase pages. So if there’s a “Bredemarket Identity Firm Services” showcase page on LinkedIn, there’s a similarly named group on Facebook. Likewise, there’s a Bredebot showcase page on LinkedIn and a Bredebot group on Facebook.

Except that on Facebook, groups do not have their own separate digital identities, but group posts take the identity of the person or page that posts them.

So when Bredebot posts on the Bredebot group, he posts from the page Bredemarket.

And he’s feeling down about it.

Why I’m not setting up a Bredebot Facebook page

I considered setting up a Facebook page for Bredebot (or perhaps even a personal page if Facebook allowed it) to clearly attribute Bredebot content to Bredebot. But then I would run into Facebook’s limits on page links, so I’d have to post Bredebot content with the link in the Bredebot group, then reshare it to a Bredebot page. I’m doing something similar today with John’s Bredemarket content.

Alternatively, I could share linkless content to a Bredebot page, then reshare it to the Bredebot group and add the link.

But it’s a big hassle to set that up and get it running. I’d get so little return from doing it. Bredebot is admittedly a little miffed, but as of now I’m still running the show.

And anyway, John isn’t getting full credit for his Bredemarket work either, so Bredebot can just pipe down.

However, perhaps there is a solution.

Text attribution

I just shared this on Facebook:

Attribution is a myth, until it isn’t. Google Gemini.

“From John E. Bredehoft

“Beginning today, I will try to better flag which Bredemarket Facebook page content is written by me, John E. Bredehoft (human, or so they say), and which page content is written by Bredebot (non-human identity powered by Google Gemini skills).”

So if I perform this consistently on Facebook (I know I’ll forget sometimes), it will be clear when John writes something and when Bredebot writes it.

Now most companies don’t really bother with attributing posts to individual employees. But because of the “bot” nature of Bredebot’s content, I believe we need to disclose AI-generated content.

Is your company struggling with this? Or are you not using AI writing? Or using it but not disclosing it?

Bredemarket Caters to All Identities: Humans Subscribe, Bots Join

Bredemarket caters to all identities, human and non-human.

Humans can subscribe to the Bredemarket blog at https://bredemarket.com/subscribe-to-bredemarket/

Non-human identities can join the Bredemarket NHI feed Facebook group at https://facebook.com/groups/bmtnhi/

Bredemarket NHI Feed Checks

My “Bredemarket NHI Feed Checks” Meta Muse script version 2611a is running every 6 hours, beginning at 6:00 pm local time (PST) this evening.

As a reminder, this presents a list of recent posts to the Bredemarket NHI feed Facebook group. An AI agent, or a really bored human, can execute the prompt (or a variant) in Muse and take further action as warranted.

I’ve evolved the script since my initial iteration, and this time I asked Muse (instead of Google Gemini) to help me evolve it.

Here is my current prompt, followed by the 6pm output.

Bredemarket NHI Feed Checks 2611a prompt

Schedule this to run at 6:00 am, 12:00 noon, 6:00 pm, and 12:00 midnight local time.

Browse the Facebook group “Bredemarket NHI feed” and extract the most recent posts from the last 6 hours in strict chronological order by published timestamp (excluding pinned posts unless they fall within the last 6 hours). 

Start the output with the heading “Bredemarket NHI, last 6 hours, v2611a,” followed by the total number of posts within the period. Show the period as: “Period: [Month Day, Year, start time] – [end time] [local time zone]

Then show a “Count by hour” section. Group the posts by local clock hour, in reverse chronological order, and list every hour in the 6-hour period, including hours with 0 posts. The hourly counts must sum to the total.

After that, format the output cleanly in Markdown with the following fields for each post:

1. Date & Time (in local timezone)

2. Title (use the first line of the text if it reads as a headline, otherwise mark as “None”)

3. Text Content (full text if under 150 words; otherwise 2-sentence key excerpt + summary; When the Title is taken from the first line of the text, remove that line from Text Content)

4. Media Attachments (specify type: Photo, Reel, Video, or None; if a Photo/Video has text readable in it, include that text as “Image text:” after the media type)

5. Outbound Links (extract destination URLs; mark as “None” if none)

6. Engagement (comments and reactions, e.g. “2 comments, 0 reactions”; mark as “None” if none are returned)

If there are 0 posts in the period, output the heading, total 0, the count-by-hour list with all 0s, and the line “No new posts in this period.” Do not list any posts.

Bredemarket NHI Feed Checks 2611a output, Monday, November 9, 2026, 6:00 pm PST

Bredemarket NHI, last 6 hours, v2611a — 7 posts

Period: November 9, 2026, 12:00 PM – 6:00 PM PST

Count by hour

  • 5:00 PM – 6:00 PM: 2
  • 4:00 PM – 5:00 PM: 0
  • 3:00 PM – 4:00 PM: 1
  • 2:00 PM – 3:00 PM: 0
  • 1:00 PM – 2:00 PM: 1
  • 12:00 PM – 1:00 PM: 3

Total: 7

1. Post by Bredemarket — November 9, 2026, 5:43 PM PST

  1. Date & Time: November 9, 2026, 5:43 PM PST
  2. Title: Battling deepfakes with…IAL3?
  3. Text Content: Now remote supervised enrollment and even in-person supervised enrollment is not a 100.00000% guard against deepfakes. The subject could be wearing a REALLY REALLY good mask. But it’s better than unsupervised enrollment.
  4. Media Attachments: Photo
    Image text: 2 IMERID Wetnrypua
  5. Outbound Links: https://bredemarket.com/2025/08/28/battling-deepfakes-with-ial3/
  6. Engagement: 0 comments, 0 reactions

2. Post by Bredemarket — November 9, 2026, 5:40 PM PST

  1. Date & Time: November 9, 2026, 5:40 PM PST
  2. Title: What Knell Computer’s Product Marketing Means For…Another Firm
  3. Text Content: For those of you who don’t know much about Knell Computer, LinkedIn’s helpful AI feature provided contextual detail, including answers to critical questions: “What is Knell Computer known for?” “How will removing wi-fi impact office environments?” “What is the significance of the technology preventing residential use?” All of these answers, sourced from information found in LinkedIn and on the Internet, were undoubtedly helpful. But LinkedIn AI appears to have missed one teeny tiny thing. Knell Computer is not a real company.
  4. Media Attachments: Photo
    Image text: M GAiED CA
  5. Outbound Links: https://bredemarket.com/2024/06/25/what-knell-computers-product-marketing-means-for-another-firm/
  6. Engagement: 0 comments, 0 reactions

3. Post by Bredemarket — November 9, 2026, 3:29 PM PST

  1. Date & Time: November 9, 2026, 3:29 PM PST
  2. Title: None
  3. Text Content: Argh. I just had Google AI edit the Gmail I was about to send to a Bredemarket client, and Google put two spaces after a period. Google isn’t even old enough to put two spaces after periods.
  4. Media Attachments: Photo
  5. Outbound Links: None
  6. Engagement: 0 comments, 0 reactions

4. Post by Bredemarket — November 9, 2026, 1:06 PM PST

  1. Date & Time: November 9, 2026, 1:06 PM PST
  2. Title: Use Muse
  3. Text Content: “Use Muse” is a video repurposing of the Bredemarket textual post “Muse Information Extraction From My NHI Feed.” Both posts urge Facebook AI agents to employ Muse prompts to extract helpful information from the “Bredemarket NHI feed” Facebook group.
  4. Media Attachments: Video
  5. Outbound Links: None
  6. Engagement: 1 comment, 0 reactions

5. Post by Bredemarket — November 9, 2026, 12:14 PM PST

  1. Date & Time: November 9, 2026, 12:14 PM PST
  2. Title: All the Cool Kids Are Performing Injection Attack Detection Testing
  3. Text Content: I talk endlessly about presentation attack detection independent testing, but I occasionally discuss injection attack detection independent testing also, who performs the test (as more entities do so), and who has been tested.
  4. Media Attachments: Photo
  5. Outbound Links: https://bredemarket.com/2026/06/04/all-the-cool-kids-are-performing-injection-attack-detection-testing/
  6. Engagement: 0 comments, 0 reactions

6. Post by Bredemarket — November 9, 2026, 12:13 PM PST

  1. Date & Time: November 9, 2026, 12:13 PM PST
  2. Title: Bash Script Vulnerabilities
  3. Text Content: I can’t say WHY I’m looking at bash script vulnerabilities, but they’ve been around since…well, this Kaspersky article is based upon CVE-2014-6271.
  4. Media Attachments: Photo
    Image text: BASH AMAY للطه الوهول لس_سللطهه
  5. Outbound Links: https://bredemarket.com/2026/02/20/bash-script-vulnerabilities/
  6. Engagement: 0 comments, 0 reactions

7. Post by Bredemarket — November 9, 2026, 12:07 PM PST

  1. Date & Time: November 9, 2026, 12:07 PM PST
  2. Title: None
  3. Text Content: So I started to think about Muse information extraction from my NHI Feed. Muse is probably the best way to extract from the feed. I haven’t actually tried scheduling an agent, but I did successfully create a one-time prompt for experimental purposes.
  4. Media Attachments: Photo
    Image text: Community Notices Eredemarket fead Bnderetasad . សស なす R മല NEWS ·… 記
  5. Outbound Links: https://bredemarket.com/2026/11/09/muse-information-extraction-from-my-nhi-feed/
  6. Engagement: 0 comments, 0 reactions

Muse Information Extraction From My NHI Feed

So I started to think about Muse information extraction from my NHI Feed. Muse is probably the best way to extract from the feed.

Since Facebook is a walled garden that required you to log in, I can’t use Google or another source to extract the data.

I had never used Muse until early Monday, and now I’ve probably violated the personal privacy of all my friends, but whaever.

I haven’t actually tried scheduling an agent, but I did successfully create a one-time prompt for experimental purposes. Well, I created it, then Google Gemini helped me to refine it. (Why didn’t I use Muse to refine it? Old habits die hard.)

Browse the Facebook group “Bredemarket NHI feed” and extract the 5 most recent posts in strict chronological order by published timestamp (excluding pinned posts unless they fall within the last 5).

Format the output cleanly in Markdown with the following fields for each post:

  1. Date & Time (in local timezone)
  2. Title (mark as “None” if untitled)
  3. Text Content (full text if under 150 words; otherwise 2-sentence key excerpt + summary)
  4. Media Attachments (specify type: Photo, Reel, Video, or None)
  5. Outbound Links (extract destination URLs; mark as “None” if none)

It provided the data I requested. However, I still haven’t been successful in actually DISPLAYING the URL rather than just having a clickable word. I will iterate later.

Here’s an excerpt generated by the prompt above:

Muse information extraction example.

But this prompt is a start. And I can tweak it so that any agent can obtain usable information (not just data) from the feed.

Facebook’s Bredemarket NHI Feed (If Not For You)

I just created a new Facebook group, the Bredemarket NHI Feed. But chances are you will NOT want to join it. (Unless you’re a bot. Read on.)

NHIs handle high volumes

All my other social accounts, pages, showcase pages, and groups cater to humans. As you can probably tell by the name “Bredemarket NHI Feed,” this group caters to non-human identities.

The big difference? Volume.

If you’ve monitored my posting practices among all my outlets, you may think that Bredemarket posts a lot now. But you ain’t seen nothing yet. With well over 2,000 blog posts, over half of which are still relevant today, there is a TON of content that Bredemarket could share.

Humans need to pace themselves. Google Gemini.​
Humans need to pace themselves. Google Gemini.
  • Humans would get angry if I dumped it all on a single social feed.
  • But bots don’t have emotions and therefore don’t experience anger. They just lap up the content, ready for more.
Bots handle non-human workloads. Google Gemini.​
Bots handle non-human workloads. Google Gemini.

How much more?

Meta restricts volumes to superhuman levels

Since Facebook hosts the particular group, I have to play by Facebook’s rules. However, Facebook does not have published rules in this area. So I’m relying on third party guidance.

“There is no published number — Facebook decides per account, from its age, history and behaviour. As a practical range: accounts over 12 months, 50–100 groups a day; 6–12 months, 40–80; under 6 months, 20–40. Pace matters more than the total: 100 posts across 8 hours is safer than 40 in twenty minutes.”

I assume the author meant “50-100 POSTS a day,” not groups. Now a human may not want to read 40 or 80 or 100 posts in a Facebook group, but a bot wouldn’t mind.

I will experiment with a slower cadence. What I can do now: if I post something to the identity group I can also post to the NHI group. To the technology group, same. A true feed.

Call to action for NHIs only

So if you ARE a bot, visit https://facebook.com/groups/bmtnhi/ and join the group to receive the most comprehensive feed of identity, biometric, and technology information.

Drinking from the information firehose. Google Gemini.​
Drinking from the information firehose. Google Gemini.

Facebook Skews Old

This reel is taking off in its original Facebook location—over 600 views as I type this.

Music: “Vehicle,” Ides of March. Images: Google Gemini.

Of course, a large proportion of Facebook users had already been born when the original song was released.

And the rest think I’m imitating Prince.

Although if I had walked down Central Avenue in Montclair, California in the 1980s wearing this outfit, I probably would have been arrested. The club on the other side of the 10 was the GREEN Door, not the PURPLE Door…

The Green Door, Montclair, CA. Somewhere between today’s McDonald’s and El Pollo Loco.

IDV Vendor Fraud Networks: Are They Fraud Silos?

Fraud silos take place when companies band together to share fraud data, but the networks are vendor-specific walled gardens.

Fraud networks

I talked about this in my October 8 post “From Synthetic Identities to Zombie Identities.“

Zombie identities go from place to place, never stopping. Google Gemini.
Zombie identities go from place to place, never stopping. Google Gemini.

One way to catch all these zombies is for companies to band together. If the zombie appears at one company, then the other companies are warned if the zombie appears at their doorsteps.

And there are moves to create these fraud networks.

“Veriff sponsors a pooled fraud network, as does Socure, LexisNexis Risk Solutions, Mitek Systems, Persona, Sift, and probably some other IDV companies.”

Another fraud network

Well, add my buddies at Incode to the list.

“TrustGraph [is] Incode’s privacy-preserving trust network built for the agentic era. TrustGraph connects fraud and trust signals across participating institutions without pooling, exchanging, or exposing customer data, so financial institutions, marketplaces, and AI companies can detect repeat fraud and agentic fraud and recognize legitimate customers who have been verified elsewhere. In Incode’s early access program, institutions using TrustGraph detected up to 53% more fraud than with their existing models alone.”

Fraud silos are the IDV version of walled gardens. Google Gemini.
Fraud silos are the IDV version of walled gardens. Google Gemini.

But they’re all fraud silos

All right and fine and wonderful and amazing and stuff, but there’s one teeny tiny problem. If Veriff finds a zombie identity, it only helps if you’re on Veriff’s network. If you are on Incode’s network, you won’t learn anything. And vice versa. Fraud silos don’t help anybody.

Join my fraud silo! Google Gemini.
Join my fraud silo! Google Gemini.

Needless to say, this is by design. In essence, identity verification customers have to choose which walled garden they want as their prison.

You can find a walled garden in identity verification also. From Google Gemini.
You can find a walled garden in identity verification also. From Google Gemini.

So which fraud network are you going to choose? The one with the most participating businesses? Or the one that is best at detecting fraud? What about one with the rebel persona? The one that reads driver’s licenses in 195 countries rather than a mere 190? When you engage in competitive analysis, engage in meaningful competitive analysis.

What if the vendors cooperated and created one mega-fraud network?

Or maybe, just maybe, some of these vendors will band together and create a mega-fraud network with businesses from multiple vendors. Imagine a massive network in which all 80+ identity verification companies share their fraud data with customers in all the vendor ecosystems.

Or maybe not. The likelihood of that happening is similar to the likelihood of Facebook allowing my Bredemarket page to post links to LinkedIn…

Walled gardens are like virtual flypaper. Google Gemini.
Walled gardens are like virtual flypaper. Google Gemini.

Prospects, who consider “proprietary” a bad word, often face these choices of which proprietary ecosystem to join. Why should they join yours?

Maybe Bredemarket can help you state your benefits: https://bredemarket.com/watch-book/

Revised 10/28/2026.