# Meta Unwinds $2B Manus Acquisition, and I had a baby!

An episode of [AI Chat: AI News on Claude, ChatGPT, Gemini, Grok, Perplexity](https://invisiblenetwork.ai/podcast/1669990482), 16 Jun 2026, 17 min.

Audio: https://content.rss.com/episodes/365073/2919963/ai-chat-podcast/2026_06_16_16_27_25_2da6e365-5866-4eea-bba7-ddfcd3f9639a.mp3

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! In this episode, we share exciting personal news alongside discussions on Meta's unwinding of their $2 billion Manus acquisition and the repercussions stemming from that decision. We also cover recent developments involving Anthropic’s AI models, Apple's Siri updates, and Jeff Bezos' Prometheus company raising $14 billion for physical AI innovation. Chapters 00:00 Introduction 00:59 Personal News 02:01 Meta's Manus Acquisition Unwinding 06:00 Anthropic's AI Model Controversy 14:59 Siri's Future Changes 16:50 Bezos' Prometheus Funding Show Links • Get the top 80+ AI Models for $8.99 at AI Box: https://aibox.ai • How I Grow and Scale My Business with AI: https://www.skool.com/aihustle • Get the AI Chat Daily Newsletter: https://www.aichatdaily.com/newsletter

## Transcript

[0:02] Welcome to the AI Chat Podcast. I'm your host,
[0:04] Jaden Schaefer. Every day, I cover cutting -edge
[0:07] AI news and talk with the leaders behind it,
[0:09] breaking down what it means for your life and
[0:11] business. Welcome to the podcast today. I have
[0:17] some exciting news. And aside from the fact that
[0:20] Meta is unwinding their $2 billion Manus acquisition,
[0:23] I have some even more exciting personal news,
[0:25] and that is that last week my wife and I welcomed
[0:27] our newest member of the family. We had a healthy
[0:31] baby boy. His name is Clay, and it is... absolutely
[0:35] exciting in our household so if you're wondering
[0:37] why my podcast posting hasn't been super consistent
[0:40] over the last week and maybe for the next week
[0:42] or two it might be a little spotty i'll do my
[0:44] best but uh i'm spending a lot of time snuggling
[0:46] that little guy currently him and mom are sleeping
[0:48] so i figured this would be a good time to get
[0:50] a podcast episode out but everything went incredibly
[0:52] well with the birth everyone is healthy and happy
[0:55] and doing great and growing he's pooping lots
[0:58] so we're pretty happy he's now five days old
[1:01] this year i turned 30 and this is our fourth
[1:04] kid so we definitely have our hands full but
[1:06] we feel super blessed I always tell everyone
[1:08] to have as many kids as you can as soon as possible.
[1:11] They are the best thing about life. Okay, let's
[1:14] talk about what's going on in the news. I mentioned
[1:16] Meta unwinding this $2 billion Manus acquisition.
[1:18] This is because Beijing told them that they had
[1:21] to divest it. By this point, we've all seen the
[1:23] headlines that Anthropix Fable 5 got banned over
[1:27] the weekend and pulled so you no longer get access
[1:29] to that. But it turns out that an Amazon security
[1:32] paper is what triggered the White House banning
[1:34] this. We'll get into that. Apple's Craig Fredger.
[1:37] says that the new Siri is going to refuse romantic
[1:40] roleplay. So we're going to talk a little bit
[1:41] about where Siri is going in the future and what
[1:44] it can and can't be used for. Jeff Bezos's Prometheus
[1:47] company has just raised $14 billion at a $41
[1:50] billion valuation for physical AI. And there's
[1:52] some interesting things being built there. And
[1:55] the White House says they suspect a China -linked
[1:57] group accessed Anthropic's Mythos model. So this
[2:00] isn't even their Fable 5. This is their Mythos
[2:02] model. And so there's some other drama falling
[2:04] out from there. The first thing we want wanted
[2:06] to get into is the meta story. This is absolutely
[2:09] crazy. I mean, this was a $2 billion acquisition.
[2:12] And what was interesting, just for like two seconds
[2:14] of context on the story, basically, it was started
[2:17] in China. It had a completely different name.
[2:19] It was like Butterfly or something like that.
[2:21] And then they moved it to Singapore. And I believe
[2:23] the CEO and the CTO were still stuck in China,
[2:27] although they moved the whole team to Singapore
[2:29] and then changed the name. And it kind of scaled,
[2:33] became incredibly popular. Meta acquired it because
[2:35] they can acquire a company out of Singapore.
[2:37] But I think because the CEO and the CTO were
[2:39] still like locked down in China, like they were
[2:41] banned from leaving. This is kind of how they
[2:44] got them to unwind the deal by sort of holding
[2:47] them. It's interesting because in this particular
[2:49] case, you know, when they first came out with
[2:52] the news telling Meta to unwind the acquisition,
[2:54] Meta said, you know, we believe we have legally
[2:57] acquired it. We're, you know, we're allowed to
[2:59] do this. We're going to, you know, discuss this
[3:02] in court with them. And, you know, Meta is a
[3:05] company that is banned in China, like Instagram,
[3:08] Facebook, WhatsApp, all of their different social
[3:12] media platforms are banned in China because China
[3:13] censors, you know, the Internet and stuff. It's
[3:16] interesting that they still are unwinding this
[3:18] deal, even though, you know, Beijing doesn't
[3:21] really have that much. You would assume they
[3:22] don't have that much sway over meta, right? Like
[3:24] they can't really tell meta what to do when they
[3:27] can't blackmail them like, hey, we're going to
[3:29] ban your other apps like they don't have as much
[3:30] leverage. So it's interesting to me that they're
[3:32] still unwinding this deal. I'm assuming it probably
[3:34] has something to do with kind of holding the
[3:35] other co -founders hostage over there and probably
[3:38] some other things. The co -founders are in preliminary
[3:42] talks to raise about a billion dollars to buy
[3:45] the company back from Meta. So I guess this is
[3:47] the way that they want to do it is, you know,
[3:50] Meta bought this for two billion dollars. But
[3:52] I think China's like, OK, fine, look, we'll we'll
[3:56] we'll basically let them raise money elsewhere
[3:58] and pay back Meta. So I'm curious to see how
[4:00] that goes. Although from what I'm seeing, they're
[4:03] in preliminary talks to raise a billion dollars
[4:05] and Meta bought it for two billion. So I don't
[4:07] know. I wonder if Meta is going to just. up lose
[4:10] money or how this whole thing will work. Meta
[4:12] has cut Manus off from their internal systems.
[4:15] They're blocking employees from using Manus tools.
[4:17] Now, what's interesting is I actually have a
[4:19] friend that works in advertising and he works
[4:22] in a marketing firm that does Facebook ads. And
[4:24] he was telling me that he loves Manus. It does
[4:27] a lot of stuff inside of his company. It's been
[4:29] super useful. And so it's interesting to me that
[4:32] once this is getting rolled back, he's going
[4:34] to get cut off from that. Now, what I will say
[4:36] is I feel like this actually Isn't that... terrible
[4:40] of a thing for meta and i think it is a pretty
[4:42] terrible thing for manis in my humble opinion
[4:44] because um claude co -work you know even my friend
[4:47] that works doing a lot of stuff inside of facebook
[4:49] ads he's like yeah like i use manis for like
[4:51] you know these three things because it has access
[4:53] to these three files but then i get i just plug
[4:55] it into co -work and it gives me like more stuff
[4:57] and it actually does more of the doing for me
[5:00] so what's interesting is i think co -work is
[5:02] much more powerful i think meta could build a
[5:04] manis clone or co -work clone themselves they
[5:07] could build that tool I'm not sure if they will
[5:10] because I'm not sure what their strategy is.
[5:12] That's kind of just a more general purpose, whereas
[5:14] they kind of bought something that they could
[5:15] plug straight into Facebook ads and they kind
[5:17] of had a good use case there. So I'll be curious
[5:19] to see what they do, although Manus was, you
[5:21] know, for kind of general purpose as well. So
[5:23] I'll be curious to see if Facebook rolls out
[5:25] their own version. Assuming they get all of their
[5:28] money back. I mean, it was a big headache, but
[5:30] I made them look cool for a minute. So assuming
[5:32] they get all of their money back, I think they
[5:34] could spend that money and build something similar
[5:36] that would be quite valuable to their company.
[5:38] anyways um so i think in my opinion bad move
[5:42] on china's part and i mean i get it they're like
[5:45] look we want to we're kind of behind an ai race
[5:47] we want to be able to hold any ai companies that
[5:50] are breakouts uh inside of china and keep controlling
[5:53] them but uh yeah i'm not sure that's it was the
[5:55] the best move but from china's side i i get it
[5:59] right it's i feel like they'd lose power to to
[6:02] the West by giving up one of their very few popular
[6:06] AI companies. Not few, I'm sure they have plenty
[6:10] of... big AI companies that are used by Chinese
[6:12] people just haven't made a lot of breakouts in
[6:14] the broader market. DeepSeek isn't really in
[6:16] the headlines like it used to be. Okay, let's
[6:18] talk about Amazon. They had a security paper
[6:21] that triggered the White House's ban on Anthropic's
[6:23] Fable model. This is interesting because it's
[6:24] a counter to the narrative that I heard from
[6:26] a bunch of people, which is basically the White
[6:27] House is mad at Anthropic, so that's why they're
[6:30] kind of banning the Fable thing. I didn't really,
[6:31] Fable 5. anthropics latest and greatest model.
[6:35] What's interesting about fable five, it is the
[6:36] mythos models. This is the model that, you know,
[6:39] they went around telling everyone this is like
[6:41] super dangerous, it can hack and breach everything
[6:42] in the world. And then they're like, hey, we
[6:44] put some guardrails on it. And now we're releasing
[6:46] it to the public, everyone can have it, but it
[6:48] just has guardrails on it. And there's a bunch
[6:49] of articles I read on AI chat daily .com that
[6:52] were basically saying, if you ask it any biology
[6:54] related questions, it would It would not answer
[6:57] them. And Anthropic said that's on purpose because
[6:59] it didn't want people to make bioweapons with
[7:01] them. So then it would forward you to Opus 4
[7:03] .8. If you ask any cybersecurity related questions,
[7:07] it would also not answer and forward you to Opus
[7:09] 4 .8 for, you know, doing any sort of security
[7:11] audits because they don't want it to find any
[7:13] codes. So it had these guardrails. They release
[7:15] it to everybody. And apparently. Amazon CEO Andy
[7:19] Jassy briefed the White House on a security report
[7:21] about Anthropix Fable 5. And basically because
[7:25] of that, they were able, they pulled, they banned
[7:29] Fable 5. And according to David Sachs, who's
[7:32] posted on Twitter, used to be the AI czar, he
[7:34] said, talking to people inside of the administration,
[7:36] because he's no longer the AI czar, but I'm sure
[7:38] he's just got a lot of the connections there.
[7:39] He said from his understanding, Amazon came.
[7:43] out with this paper, said they had, you know,
[7:45] figured out ways to get around the guardrails.
[7:46] They told Anthropic to fix them. Anthropic refused
[7:48] to fix them because they said it wasn't necessary.
[7:50] And then the U .S. government said, OK, well,
[7:52] in that case, we'll ban it. I'm sure Fable 5
[7:55] is going to come out again. Hopefully that will
[7:56] get fixed. David Sachs said he's really perplexed
[7:59] because. He loves Anthropic. He loves what they're
[8:03] building. And he said the administration has
[8:05] a lot of respect for their technical capabilities
[8:08] of their models. And to be honest, it felt like
[8:10] the whole relationship there was kind of thawing
[8:12] out because they were briefing the different
[8:14] departments of the government. So it felt like
[8:15] it was kind of thawing out, which is what I personally
[8:17] would like to see more of. But he said he was
[8:20] perplexed that they didn't want to go and. update
[8:23] those safety models. Maybe it was hard. Maybe
[8:25] I'm not really sure what's going on. But this,
[8:27] this will be interesting. Anthropic disputed
[8:30] the jailbreak, apparently the one that Amazon
[8:32] came out with. And they said they argued basically
[8:35] the same prompts and outputs work on GPT 5 .5
[8:38] and any other public models. So, you know, kind
[8:42] of what they're saying and the reason why they
[8:43] didn't fix quote unquote fix is because they're
[8:46] like, look, yeah, you can get around some of
[8:47] our guardrails with some of those techniques,
[8:49] but you can use those same techniques to go quote
[8:51] unquote jailbreak GPT 5 .5. So why do we have
[8:53] to? fix it, right? So it is a fair point, right?
[8:57] Is GPT 5 .5 going to get banned or, you know,
[9:01] get export controls put on them, etc. So it,
[9:04] you know, people could say it's one sided, but
[9:06] at first a company that's so focused on being
[9:08] a safety company, like that's basically Anthropix,
[9:10] you know, mantra, I would assume they would take
[9:13] a slightly different approach and, and try to
[9:15] work with people to fix it and not sweep it under
[9:19] the rug, mostly because they just came out and
[9:21] did their big PR stunt saying, you know, Mythos
[9:23] is gonna to take over the world, they got a lot
[9:24] of people that aren't even that into AI concerned
[9:26] and know about that story. So I would think they
[9:29] would, you know, for nothing else, for that reason,
[9:31] take it a little bit more seriously. Okay, Apple's
[9:33] Craig Frederick said that the new Siri is going
[9:37] to refuse romantic roleplay. This is interesting
[9:40] because Apple has kind of been on the sidelines
[9:42] with a lot of this AI for quite a while, and
[9:44] they're finally coming out with their updated
[9:46] Siri. And so we don't really know all of its
[9:48] capabilities, what it's going to be capable of.
[9:50] We know that you can plug ChatGP, Gemini, Anthropic,
[9:52] different models into it to power it. But it's
[9:56] interesting because this is one of the first
[9:57] times I'm seeing they're making some decisions
[9:59] on what it will do. It's not just like, look,
[10:02] plug these other models in and chat with them
[10:05] like you would chat with ChatGPT. It looks like
[10:07] they're actually setting restrictions and rules
[10:10] on the model and probably for, you know, maybe
[10:14] like safety reasons or all sorts of different
[10:16] reasons, but they have like their own controls
[10:18] that they're going to be putting on top of the
[10:20] model. OpenAI and Google's chatbot. According
[10:24] to the person over at Apple, Frederick Gray said
[10:28] that they are tuned for sycophancy and user stickiness,
[10:31] which, to be fair, if you ask a question, it's
[10:33] going to answer. And then it's like, would you
[10:35] like me to do X, Y, Z, X, Y, Z and X, Y, Z? Like
[10:37] it's trying to prod you on to keep chatting with
[10:39] it. It wants to, you know, maximum conversations.
[10:42] So that was his point. And he framed Siri's role
[10:45] a lot more narrowly. He said its goal is to answer
[10:49] the question, complete the task. It's going to
[10:51] decline the emotional bond. And it's basically.
[10:53] opposite of how a lot of the main AI apps are
[10:56] working right now. OpenAI apparently has loosened
[10:58] some adult content restrictions on ChatGPT. I
[11:00] remember Sam Altman making a big announcement
[11:02] about that. He walked it back when they kind
[11:04] of axed the Sora video model. And I think the
[11:07] whole... adult content was something that kind
[11:11] of got cut back as well. So I'm not sure where
[11:12] they are on that as well. Apple is trying to
[11:16] basically differentiate and they're going to
[11:17] put these new guardrails on that stops any of
[11:19] that kind of, and they're just saying like, look,
[11:21] that's not the use case. They said maybe they
[11:25] were going to be more of the quote unquote boring
[11:26] assistant, but it's just going to be useful privacy
[11:28] on device. And I mean, For me personally, I'm
[11:32] happy if there is a model like that that is useful
[11:35] for my kids. Like my kids could go and talk to
[11:37] Siri and use it versus me worrying about, oh,
[11:40] shoot, like if they're talking to chat, is it
[11:42] going to say something weird or something weird
[11:43] going to get triggered or, you know, who knows
[11:45] what rabbit hole it goes down. So I do see a
[11:48] lot of value to making some of those editorial
[11:51] decisions. Okay. Jeff Bezos, his project Prometheus,
[11:54] his company Prometheus, just raised $12 billion
[11:56] at a $41 billion valuation. He's working in the
[11:59] physical AI space. This is their Series B that
[12:02] they're doing. Total funding they've raised is
[12:04] $18 billion, and they've done that in less than
[12:05] a year. It is building software to automate the
[12:08] design and manufacturing of really complex hardware.
[12:11] So they're doing things like jet engines. They're
[12:13] also doing drugs. And this is basically a category
[12:16] that a lot of investors right now are seeing
[12:18] as more defensible than just software, right?
[12:21] Anthropic is going... which is just software,
[12:23] but it's incredible. And so they're trying to
[12:25] get into, you know, how do we actually manufacture
[12:27] some of these harder things? You know, Anthropix
[12:30] helping you build an app. Well, they'll help
[12:31] you build a jet engine. So it is pretty interesting.
[12:34] JP Morgan, Chase, Goldman Sachs, BlackRock, all
[12:36] of them joined in this round. They have about
[12:39] 150 employees in San Francisco, London, and Zurich.
[12:42] And a huge chunk of the $12 billion is going
[12:44] to fund compute infrastructure, right? It's so
[12:46] interesting when these companies are raising
[12:47] these massive rounds. It's not like they just
[12:49] raise them to increase salaries, although it's...
[12:51] It seems like Meta was paying some pretty huge
[12:53] salaries a while back. They were on a bit of
[12:55] a tear there. But it feels like they're just
[12:57] raising all this money because they need to buy
[12:58] GPUs. They need to buy data centers. They need
[13:00] to buy all of the infrastructure. So Prometheus
[13:02] cited Google's $920 million monthly SpaceX bill
[13:06] as kind of a floor for some serious physical
[13:09] AI programs. Bezos is arguing that AI is going
[13:13] to create labor scarcity rather than mass job
[13:15] loss. It's kind of a different narrative than
[13:18] a lot of his other peers who... have forecasted
[13:21] a huge white -collar displacement over the next
[13:23] few years. Project Prometheus right now, they
[13:26] have a $41 billion valuation. I think this is
[13:29] basically showing a lot of people believe a lot
[13:31] of capital is flowing into this area where you're
[13:33] focusing on hardware, you're working with data,
[13:36] you're working with manufacturing, there's regulatory
[13:37] moats. It's not just pure software, helping people
[13:41] build software play like you have something with
[13:43] Anthropic. The White House has suspected that
[13:47] a China -linked group got access to Anthropic's
[13:49] Mythos model. Now, this is kind of funny coming
[13:51] on the back of everything with Fable 5 getting
[13:54] banned, but Mythos, right, very hyped, very scary
[13:57] cybersecurity model. Anthropic gave it to a bunch
[13:59] of trusted partners. They gave it to Amazon.
[14:01] They gave it to Microsoft. They gave it to Mozilla.
[14:03] They gave it to Apple. A lot of their quote -unquote
[14:05] trusted partners, they said, hey, look, we found
[14:06] security vulnerabilities in all your stuff. Take
[14:08] this, run it on it, and fix all your security
[14:10] vulnerabilities before it releases to the public
[14:11] so hackers aren't going to be able to get a lot
[14:14] of it. the legacy databases they gave to a lot
[14:15] of open source projects that were used by a lot
[14:17] of people. Now, apparently, of all of the quote
[14:20] unquote trusted people they gave it to, some
[14:23] Chinese linked groups did get access to it, according
[14:26] to the White House, who now has export controls
[14:29] on mythoses on the mythos model. And because
[14:33] of this, according to a semaphore report, the
[14:36] breach they said would is basically a critical
[14:39] vulnerability where they're showing that a rival
[14:41] state with query access to this model can use
[14:44] distillation to train smaller capable copies
[14:46] without needing the outright original weights.
[14:49] Now, basically, what this is saying with distillation
[14:51] is you can take a model like chat GPT, and you
[14:53] can ask it 100 ,000 questions and see how it
[14:56] responds to those questions. You could take all
[14:58] of your questions and its responses and train
[15:00] a model to answer like those 100 ,000 questions.
[15:03] And you can actually get it's called distillation,
[15:04] but you can get a pretty good copy of that model
[15:07] and how it interacts, you give it as much real
[15:09] world knowledge outside of that as you can. But
[15:12] basically, it's like a lot of the pre training,
[15:14] this is how you it's kind of crazy. It's like
[15:16] there's less IP for these AI companies, because
[15:19] you can do this distillation method. Now, whatever,
[15:21] a lot of people have been doing this, I think
[15:23] Elon even admitted that Grok did distillation
[15:25] based off of open AI. Now, what's interesting
[15:29] is if a Chinese group got access to Mythos, the
[15:31] security one, and they were able to do this,
[15:32] they could technically create smaller, more like
[15:36] smaller capable hacking models that are good
[15:39] at different things or, you know, all sorts of
[15:42] adversarial tactics. And even if Mythos gets
[15:45] shut down, if you got access and get 100 ,000
[15:47] queries in, you could kind of distill these smaller
[15:50] models. So there's a Discord group that was previously
[15:53] talking about how they got access to Mythos for
[15:57] two weeks. weeks before Anthropic detected and
[15:58] shut down this breach. And I think that they
[16:02] were like on the website. And anyways, they figured
[16:04] out how to do some stuff. And so Anthropic told
[16:07] Semaphore that the government did not mention
[16:09] China during their export control talks. So this
[16:12] is kind of... contradicting a lot of the reported
[16:14] China motivation restrictions. And I think it
[16:16] creates a bit of a gap between what we know and
[16:18] what the policy is. So this will be something
[16:21] that has to be worked out. But I mean, if it
[16:25] really is a massive risk, you definitely don't
[16:26] want it to be leaked. Thank you everyone so much
[16:29] for tuning into the podcast today. If you enjoyed
[16:31] the episode, please leave a rating review wherever
[16:32] you get your podcasts. I'll try to keep up to
[16:36] date on my upload schedule, although I do obviously
[16:39] have the new baby. So lots of fun with him. over
[16:42] there. I was up clod coding between the hours
[16:45] of 1am and 3am last night while rocking him on
[16:48] a bouncy ball because he was a little fussy and
[16:51] was trying to let mom sleep a bit. So lots of
[16:53] fun over here, but I'm getting a lot of clod
[16:55] coding done. Anyways, hope you guys all have
[16:57] a fantastic rest of your day. Make sure to go
[16:59] check out AIbox .ai if you want to get access
[17:01] to all of the latest AI models in one place.
[17:04] I'll leave a link in the description. Catch you
[17:05] guys all in the next episode.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.