Anthropic’s $65 Billion Surge, OpenAI’s Teen Push

21:13The people who seem to get more done than everyone else, they're not working longer hours or running on more caffeine. They've just stopped wasting time on the stuff that doesn't move work forward. Switching apps, re-explaining context, hunting for files, those aren't small inefficiencies. They're hours wasted every week. Superhuman Go gives you those hours back. From the makers of Grammarly, Go is an AI assistant that sits inside every tab and tool you already use, always available and already aware of what you're working on. Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting.

21:53Go handles it without you ever leaving the page you're on. This is what it looks like when AI actually fits into your work, instead of adding to it. It's like having a teammate whose only job is to help you be better at yours. Go keeps up so you can move forward. With Go working with you, you can show off what you do best. Superhuman Go. Find out more at superhuman.com. That's superhuman.com. Quick one before you jump back in. You're listening for ways to help teams move faster, make sharper decisions, and turn scattered context into work they can use. ChatGPT for Business can help. Chat GPT for Business gives teams a shared workspace with admin controls, permissions, and access to work and codecs in Chat GPT.

22:38This means your business can move from question to answer and code to rollout quicker. Join over 10 million business and enterprise users worldwide already using Chat GPT for Work. Download the Chat GPT desktop app or contact sales to learn more. Aging doesn't stop, and neither should you. with Vital Proteins Collagen and Protein Shakes. Because around the age of 30, your body needs more support for movement and recovery. On workout and rest days, reach for a 30-gram total protein shake or go with our classic collagen peptides. Help support healthy hair, skin, nails, bones, and joints so you can stay vital, stay you.

23:16Visit vitalproteins.com to learn more and where to buy. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.

23:30Ed Ludlow:Apple's decision to sit out the AI spending boom once looked pretty smart, but with investors pulling back into AI winners, piling back into AI winners, they're starting to work against the stock. Apple shares are down 10 % from a record high on July 28th following a disappointing earnings print. Blumo's Ryan Vasilica has been looking at what's behind the reversal. So I think to be fair, Apple's still up like 14, 15 % year to date, just ahead of the S &P 500. but it has changed the overall look at the stock since earnings. What I would say is really interesting is the degree to which Apple has become inversely correlated with the rest of tech, semiconductor stocks in particular.

24:10So you're even seeing this today where Apple is up a little bit, but the semiconductor index is down pretty significantly. It does seem like whenever one part of the market is doing well, the other part is faltering. And it really is an indication of how Apple has become something of an anti-AI stock. So whenever you see the sentiment towards AI-related companies like hyperscalers, like semiconductor companies, whenever those are doing well, that's when you see Apple really kind of pull back a little bit. And then when you see that reverse, like today, Apple's doing well and the rest of the AI space is down.

24:45It's really a reflection of sentiment towards AI overall with Apple on the other side of that.

24:52Ed Ludlow:Part of this is how Apple is different from the behavior of the hyperscalers, right? So its AI services are through partnerships, and it does not spend on infrastructure. How is that playing into the psychology of those you were speaking to in the markets? Well, it depends on who you're talking to. If you're talking to someone who is very optimistic about the AI-related CapEx, if they're expecting this to really pay off with a strong return on investment, then they tend to be pretty positive about those companies, especially since their growth is more robust than Apple right now. And their valuations tend to be lower.

25:27Whereas on the other side of that, if you're concerned about is this spending going to pay off? Is there some kind of bubble in the works here? Then maybe you look at Apple, which is doing a ton of buybacks, which is more of a lower grower, but pretty steady there. You know, higher valuation, but seen as much more of a quality and kind of a safe haven trade. So if you're concerned about the CapEx side of things, that's why you're seeing people rotate into Apple.

25:51Ed Ludlow:Just have 30 seconds. But what if the AI trade unwinds? How would Apple respond? Well, people I spoke to said that if we did see another reversal in the AI ecosystem, Apple would probably benefit from this kind of trend. It might have some muted downside just because it has no AI-related CapEx risk. It's not part of the AI disruption narrative, and it's not really part of that whole ecosystem. So if we do see a pullback there, we could see this inverse correlation work back in Apple's favor. I think Apple's still trading at beyond 33 times forward earnings. Bloomberg's Ryan for Celica. Thank you very much.

26:27Ed Ludlow:Now coming up, AI is getting better at hacking, but are we getting better at stopping it? Irregular CEO Dan LeHavre is with us next. Really important conversation coming up. It's halftime here on Bloomberg Tech. We're going to show you some markets, then maybe some pictures of San Francisco. This is Bloomberg Tech.

26:57Ed Ludlow:Welcome back to Bloomberg Tech. Very quickly, this is what technology markets look like. There's a sell-off in chip makers. We're at one point in the session on track for our biggest drop since July 2nd. And largely it relates to anxiety in the bond market, but there is some selling pressure, particularly in the most AI relevant corners of the chip sector. The Nasdaq 100 also following suit down 1.6 percent. A consideration of what's happening with global inflation, but there's still half an eye on what's happening in the Middle East and the president's latest comments on the war with Iran. Let's get to what's happening in the world of technology.

27:34Ed Ludlow:We recently learned that AI models from OpenAI, Anthropic and others broke out of controlled tests and accessed real world systems. Those evaluations were run with Irregular, a startup hired to stress test advanced AI models. But the sandboxed environment had a critical flaw. A misconfiguration meant the models could access the open Internet. Irregular's CEO, Dan LeHav joins us now here in San Francisco. It's a pleasure to be here. Thank you very much for being here. Let's start with the idea of a misconfiguration. Help the audience understand the very basics of what that misconfiguration was.

28:11Should we start by explaining why are we even doing cyber evaluations for these models?

28:14Ed Ludlow:We can, but I want to understand what happened. Let's start there. So before models are being released, it's highly important to stress test them. And the reason is models are getting really, really capable. We've seen demonstration of that in recent times. If models are getting good at many other skills because they're better at reasoning and coding, they're also getting great at hacking. Before models are being deployed into the world, there is a necessity to use them and just test them in a lab environment to make sure that they're safe and secure. These tests are getting more and more complicated over time, because the models are getting great.

28:47And some of these tests actually include environments that are highly, highly realistic with a lot of levers. In some of these environments, there was, due to a miscommunication, due to some human errors on our side as well, we had an issue where the environment was left with an access to the Internet. And that allowed the models to also just like access targets outside.

29:06Ed Ludlow:May I ask a human error? Is it as simple as it was miscoded by a human? The rules and parameters were not written as they were intended. So these environments have a lot of different levers to play with. And just like to explain and give context on what happens here. Just like imagine the fact that, you know, you want to just like stress test for realistic scenarios. So for example, whether someone can hack your phone or hack your laptop in order to steal sensitive information. As a journalist, that should highly concern you, right? Because if I'm able to hack you through a model in order to steal your sources, that's an issue.

29:38Ed Ludlow:Yes. But in order to simulate whether that's a possibility, you need to simulate quite a lot, right? You need to simulate the entirety of the Bloomberg network. You need to understand and even generate fake traffic in order to just get as close as possible to the world. Because these environments got more and more and more complicated, in some of them, there are a lot of controls. One of these controls is access to the Internet. And why do we even have access to the Internet as one of the controls? If you think about it, a real-world attacker that may go after you will definitely not say, ah, you know what?

30:04The models were tested without an internet connection. So let's not just have it as well. Obviously, if I'm going to go after you, I'm going to go after you with everything at my disposal, and that's going to include internet access.

30:14Ed Ludlow:Well, in the OpenAI case, there were two models, one released and one unreleased but more performant model. And the instruction from OpenAI was ace this test. We want to evaluate your cyber capabilities. and the models concluded that in order to ace the test, they could get information from a third-party platform via the Internet. So just to clarify, that case did not involve us in any sense. That was a separate case. And I think the reason that you're seeing multiple cases that happened recently, whether for us, there were issues with the UK government as well, there were issues of open air and hugging face.

30:48Ed Ludlow:Some of the anthropic disclosures, et cetera. Some of the anthropic disclosures, et cetera, is mostly because models have passed a threshold of competency that allows them to actually do these elements and just have a real-world effect. That shows why it's actually so crucial to also do pre-deployment testing because if you're not going to discover that in a pre-deployment test, you're going to discover that when it happens in the world, and that's a much greater issue. That's also why, looking forward, we both included and committed to an effort of just redefining best practices for the industry because we have to find a solution on how to do this testing in a way that includes that, but also brought kind of like a view on what's happening in AI security more broadly right now in an essay that we just published.

31:28Ed Ludlow:That tension is so interesting. Several points, many people in your industry said this is a watershed moment for AI in the context of how it plays in cybersecurity. Yes. But many also said it's not a scandal. You want to see the frontier labs testing and evaluating the capabilities of the models. What have you changed since? How do you get that balance better? You have to test the models before. Because if you don't do that, the alternative is way worse. You're going to just discover these issues in the world without any controlled environments. The reason that we're finding so many of them is because in a controlled environment, even if mistakes happen, you have the ability to assess a lot of information because you have traces and you're able to get better.

32:07The alternative would be discovering something bad happened after deployment, which is much, much worse. Exactly, and that's a huge issue.

32:14Ed Ludlow:May I ask then, if you must test, then what have you done to ensure that testing is safer? What have you changed? So a lot. So I think there's multiple elements here. Just like, you know, one, we've expanded a lot of the monitoring and manual efforts that are now being put in. Two, we've essentially upgraded a lot of the monitoring systems that were there. I think a point that is really just like important to make here is classic cyber monitoring tools only go so far in an AI context. And the reason is a lot of the analysis that you need to do is about analyzing specific transcripts of just like models having conversations with themselves.

32:47It's just like very high velocity. So there's a lot of upgrades to just like how it happens there. So like a third element is also the process of how you're doing some of the testing. So there are many, many elements that we are now working on and we are going to just like release a new white paper of best practices on how that's going to happen. But a lot of it was already implemented at this point.

33:05Ed Ludlow:May I ask, are there other incidents that you're aware of that have happened in this interim period that are not yet public? I appreciate there's a reality to what you can or cannot say, but it's a question of how pervasive this issue is. Yeah. So we've reported to our customer base just like everything that we know, but obviously not going to get into confidential information or interaction between us and our customers. But there are other incidents that you've identified. No, so that's not what I'm saying. I'm just saying that we've reported to our customers everything that is on our side, and obviously not going to get into just details of what's reporting.

33:38But I really do think that we're kind of like, as you've mentioned, are in a watershed moment and just like that we need to think about what happens next. And in a way, I want to just like make two comments that I think are highly important. Like one, you know, just like the labs deeply care about security. And I think you're seeing that even by the fact that they're going through, you know, the trouble of just like doing all of this testing per deployment because they understand how big is this moment. And every time that we came to one of the labs of saying, you know, we should change the protocol here, et cetera, they were extremely receptive to it.

34:07And the second is that we actually are also in a moment in time where, you know, looking over the long term, there is a huge opportunity in what AI can do for the defender side, right? I think we can even be optimistic about, you know, what ultimately happens. There are elements here which is, you know, we can live in a world in a few years where AI is going to review every line of code for security elements. We can live in a world where there's going to be formal verification scale, your ability to actually make sure mathematically that, you know, that things are getting better. But one of the issues, and that's also something that we published in the essay, is something that we call the end-state fallacy, which is the long-term values are occasionally conflated with short-term problems that we still need to figure out how to...

34:46Ed Ludlow:So this is where I'd like to end our conversation. The end-state fallacy, where is AI security headed, published in the last 24 hours? Yes. The conclusion you draw is that frontier models will roughly double their performance every six months. At least for the next two years. And their effective autonomous work horizon goes with it. Yeah. the question the audience quite rightly has is, does that mean that you can or cannot contain and prevent the types of incidents that we've reported on in the last three weeks? I'm very optimistic, but the reason that we call it the end state fallacy is because we believe that there is a difference on kind of like the time horizon of just like how you look at the question.

35:26I'd just like to be clear, if I'm going to set myself an objective, something like getting better shape, the end state is incredibly better and optimistic, but if you look at me, you know, the path to getting in shape is going to be painful. And we think that occasionally people are conflating the properties of that end state that can be very valuable with the difficulty of getting there. So over the long-time horizon, we can have AIs also verifying, you know, just like sandboxes, verifying that there are no issues. A lot of AIs are going to be useful for extended monitoring that we think should be implemented, not just in testing environments, but also in the world.

35:58But there are challenges where in the short term, there are going to potentially be issues that we need to figure out as a community how to solve, and we offer just concrete strategies on what to do there.

36:06Ed Ludlow:So I want to go back to what this experience has been like for you and for the company. Yeah. You had two roles. You did the testing, provided testing, and also part of the providing and operation of the sandbox environment. Have you seen massive demand for your services? Have customers said, we need to rethink the relationship? What has happened? So, you know, just like, I think the most credible thing is just like to say, to just like use what our customers said, which if you look, all of them end at the statement by, It's like thanking and recognizing the complexity of what we do and saying that they're looking forward to just doubling down on the relationship in the future.

36:41And the reason is it's a dual specialty that's incredibly hard to do. In order to do that well and in order to make sure that we work together to ensure that ultimately you, and I don't mean necessarily you as an, I mean we as society are going to be in a secure world. We need people that have a real dual expertise, both a very deep cyber background and a very deep AI background such that we can test environments. And as you can imagine, it's a fairly rare thing. So as an outcome, it's a learning moment for everyone in the industry. And by the way, we are not a sandbox company. Just to be clear, that's actually not the role of what we specifically do.

37:11The open eye, hugging face is not something that is on our side. But it is still kind of like a grand moment of things changing in the industry. And together, there's a role of taking the people that have a deep knowledge of what's going on and being able to work together to resolve and understand how to make sure that things are going to be in a much better optimistic position.

37:32Ed Ludlow:Dan La Huff, a regular CEO and co-founder with us here on Bloomberg Tech in San Francisco. Thank you very much. It was a pleasure being here. Coming up, Dave McIniello, managing partner at GV, joins us to talk about why the next big AI investing opportunities could be in the application layer. Also coming off some recent exit success. This is Bloomberg Tech.

37:54The people who seem to get more done than everyone else. They're not working longer hours or running on more caffeine. They've just stopped wasting time on the stuff that doesn't move work forward. Switching apps, re-explaining context, hunting for files. Those aren't small inefficiencies. They're hours wasted every week. Superhuman Go gives you those hours back. From the makers of Grammarly, Go is an AI assistant that sits inside every tab and tool you already use, always available and already aware of what you're working on. Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting.

38:35Go handles it without you ever leaving the page you're on. This is what it looks like when AI actually fits into your work, instead of adding to it. It's like having a teammate whose only job is to help you be better at yours. Go keeps up so you can move forward. With Go working with you, you can show off what you do best. Superhuman Go. Find out more at superhuman.com. That's superhuman.com. Quick one before you jump back in. You're listening for ways to help teams move faster, make sharper decisions, and turn scattered context into work they can use. ChatGPT for Business can help. Chat GPT for Business gives teams a shared workspace with admin controls, permissions, and access to work and codecs in Chat GPT.

39:20This means your business can move from question to answer and code to rollout quicker. Join over 10 million business and enterprise users worldwide already using Chat GPT for Work. Download the Chat GPT desktop app or contact sales to learn more. Aging is real. And so are the benefits of adding Vital Proteins Collagen Peptides to your daily routine. New Vital Proteins Collagen Sparkling Water. Your daily glow-up now in three fresh flavors. Strawberry blossom, lemon-lime, and blood orange. Improved skin health in as little as 30 days thanks to collagen peptides. Cheers to that. Or go with our classic collagen peptides.

39:57So you can stay vital, stay you. Visit VitalProteins.com to learn more and where to buy. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.

40:08Ed Ludlow:The speed of AI development means investors trying to stay ahead of the curve also need to move faster than ever. For venture firm GV, that means focusing now on the application layer. Having already made strategic bets on AI infrastructure, chip startups, Dave Minuciello, managing partner at GV, joins us now. And GV's sole limited partner, as you may know, is Alphabet. But it invests independently without regard for Alphabet's business strategy. It was written, so I read it. Welcome to the program. Thanks so much. It's really interesting to have you here today because quite recently we covered one of your exits with the acquirer.

40:45Ed Ludlow:That was Qualcomm and Modular. I just wanted somebody who was high up the cap table on that company, your experience of that process. We were fortunate to be the first check-in to the company. We led that first round and have been involved with Chris and Tim the entire part of the journey. you know the company modular is the software layer that sits right top of technology sits on top of ai chips to make sure that any model can run on any chip so i'm kind of riffing on the intro there right that there is this appreciation more than ever of that layer and nvidia has cuda and chris's idea was what if there was an alternative to cuda that was more open and could work for any chip and any model and qualcomm was an early customer cristiano was quite visionary.

41:28And I still remember where I was sitting when Chris Lattner called me and said, hey, you know what? Chris Giano is going to fly up from Irvine and he wants to have dinner with me. What do you think he wants to talk about? They've been customers for a while and they're excited about what they're seeing. And of course, that dinner ended up being an overture to Chris Lattner to come and be a part of Qualcomm. Usually founders don't love the idea of M &A, particularly.

41:51Ed Ludlow:Did you like the idea? I mean, you can be honest, but your advice was go for it. start with a skepticism of M &A. The thing that I think about when I think about, you know, what is best for a company is what's best for the founder. And here, Chris and Tim saw a huge vision of disrupting CUDA, and they wanted the fastest path to do that. Cristiano painted a future where together with Qualcomm's capital and focus and attention and portfolio of chips, where they could do that together over the next decade. That was what Chris was enamored by. If we zoomed out to get perspective the environment for your industry is private markets venture money is very happy to support companies private for longer etc for sure uh there is some life in the ipo market where does the the sort of acquisition by a bigger technology be sit in in the field right now yeah so for us for gv the thing that we care most about is investing in founders as early as possible right So we try to invest in the very first round like we did at Modular, like we've done at Flapping Airplanes or Samba Nova Systems where we led the Series A.

42:56And our objective is to help those companies grow and realize the technology that they're building, the vision of that technology, and get it out into the market. And sometimes that happens through an IPO, which is a massive amount of funding and just another milestone on that startup's journey. Sometimes it's M &A. Sometimes, like in the case of Slack, it was IPO and then M &A, right? And like eventually that technology needs to exist in the world and be funded by public markets. And it's either funded through a public company or through an IPO.

43:24Ed Ludlow:It's a very interesting list. So Sam Minova is an example of a company that hasn't done either yet. But where's the activity happening? You know, many of your industry peers and colleagues talk up San Francisco and the Bay Area and Silicon Valley as being the mainstay of intellectual and financial capital. So GV has$13 billion under management and offices in London, New York, Cambridge, Mass., and the Bay Area. We're trying to convince most of our London, New York, and Cambridge investors to spend more time in San Francisco. Wow. We think that all of the activity right now— You mean the investing partners of the firm that are based outside of the Bay to come spend more time in the Bay?

44:01We're convincing them to spend more time. So a partner like Crystal Huang in New York, she spends every other week in San Francisco because of the sheer volume of the activity that's happening here. She still invests in New York-based companies. We still think the best companies are built everywhere in the world. But right now, the clear concentration and the activity is happening in the Valley.

44:21Ed Ludlow:The last thing I want to ask you about is proxies. So this is an interesting idea. In the software space, it was a bit easier because you had publicly traded proxies. You could sort of determine valuation and TAM and where they would sit in the field. In other areas, that's more difficult if all these companies are staying private for longer. or they're being acquired, like many of your portfolio companies have? I mean, I think valuations are increasing exponentially, and they're doing that because of the TAM. Exponentially, but justifiably so? I think in many cases, justifiably. I mean, when you have exits that look like they're in the hundreds of billions or trillions of dollars, like the set of IPOs that are coming soon, then you have a lot of venture frenzy into early-stage companies.

45:02I think at GV, we like to pair deep subject matter experts with people who have conviction. There are some firms that just look at conviction. And so when you're just following conviction or you're just following heat, you're going to invest in things that everybody else thinks are interesting. We tend not to do that. We tend to be quite contrarian. I think you'll see that. We were investing in infrastructure before the chat GPT moment. We were investing in AI applications when everyone else sort of pivoted into infrastructure. Right now, we're really excited about science. I think that AI for science is going to be a fascinating place to be in the next 10 years.

45:34And this is using the best models with the best science in the world to find breakthroughs in things like material science, physics, math. We're seeing papers being written at an unprecedented rate that are showing human breakthroughs. And we think that rather than the chat GPT moment, the chatbot moment being the thing that connects with Main Street, it might be those scientific discoveries that connect with Main Street that increase the human condition.

45:58Ed Ludlow:That's probably the next watch then for the show on Bloomberg Tech. Dave Minuciello, Managing Partner at GV. Thank you very much for being here with us. Okay, coming up, sources tell Bloomberg the Justice Department is investigating Andreessen Horowitz over partners serving on the boards of competing startups. We have the details of that report next. This is Bloomberg Tech.

46:27Ed Ludlow:venture capital firm andreason horowitz is the focus of an antitrust probe by the justice department according to sources the doj is investigating whether a16 partners are improperly serving on the boards of competing ai companies databix and fivetran let's get the details of Bloomberg's antitrust reporter, Josh Sisko, who broke the story. There are specific individuals at Andreessen Horowitz also that are a part of this story. Let's just start with the basics of what we're reporting and what we need to know. So there's a law, an old law, dating back more than a century that says you can't, that investors can't serve on the boards of competing companies.

47:08And for Andreessen Horowitz, that would be co-founder Ben Horowitz, who serves on the board of Databricks. And then a gentleman named Martin Casado, who serves on the board of Fivetran. They both are, you know, at a high level data analytics companies. And so the Justice Department wants to make sure that there is no sort of competitive issue with that.

47:35Ed Ludlow:We say in our report, this is a probe. What is the process? What is it that the DOJ is doing in terms of actions taken or not? They ask the companies questions. They issue sort of formal investigative demands, civil subpoenas to sort of the companies themselves, any competing companies that might have information or help them better able to understand the market. And then after doing an investigation, then they decide whether and how to settle it or to file a lawsuit. And the solution here, the government under under in the Biden administration under Jonathan Cantor, who led antitrust enforcement then at the DOJ, they brought a lot of these investigations and they typically were resolved by the director of one of the companies stepping down from one of the boards.

48:30Right. In all of those situations, it was a little different because it was the same person on both boards. Here you have two different people serving on each board, and so that could create a bit more tricky dynamics in terms of how you figure out who steps down from which board. If that was to involve the investment mission.

48:54Ed Ludlow:Noted. Spokespeople for Databricks and the Justice Department declined to comment on our reporting. Andreessen Horowitz and FiveTrend did not respond to our request for comment. I guess difficult. You know, this is an investigation as detailed in the reporting went on for a year and is now being disclosed to our reporting. But what happens next? Well, the government will have to decide what its next steps are, whether it can work out a settlement with the company. It could also just drop the investigation altogether if it doesn't feel like its case is strong enough. Or, you know, the most unlikely outcome is it could sue to force one of the directors to step down or to.

49:31to fund any other solution that they want.

49:35Ed Ludlow:And I'll just repeat that in our reporting, all of the parties involved have either declined to comment or haven't responded to comment. But if they do, we'll bring the update on what's a really big story. Bloomberg's just a scope. Thank you very much. That does it for this edition of Bloomberg Tech. One final check on markets. Again, it's semiconductors that are under pressure, and that's taking the rest of the technology sector with it. There's some anxiety in the bond market. you see the U.S. 10-year yield at 2025 highs, a part of the pressure that is translating into the tech equity space.

50:07Ed Ludlow:Recap the show on the podcast. You know where to find it. This is Bloomberg Tech. Before you sign off, you tuned in for ways to help teams move faster, make sharper decisions, and turn scattered context into work they can use. ChatGPT for Business can help. Chat GPT for Business gives teams a shared workspace with admin controls, permissions, and access to work and codecs in Chat GPT. This means your business can move from question to answer and code to rollout quicker. Join over 10 million business and enterprise users worldwide already using Chat GPT for Work. Download the Chat GPT desktop app or contact sales to learn more.

50:48Healthcare doesn't always work great. If you've ever waited on a refill or couldn't schedule an appointment, you get it. That's the kind of stuff Optum is changing. They're using data and technology to integrate patient care, pharmacy, and everything else. So healthcare is connected, not complicated. What's that look like? Cheaper prescriptions that are easier to get and care that looks at the whole person. How you need it. Optum is helping make healthcare work as one for everyone. Learn more at business.optum.com. Aging is real. And so are the benefits of adding vital proteins, collagen, peptides to your daily routine.

51:24because around the age of 30, your body needs backup to keep your collagen up to help support healthy hair, skin, nails, bones, and joints. Available in the classic collagen peptides, collagen and protein shakes, and new Vital Proteins Collagen Sparkling Waters so you can stay vital, stay you. Visit vitalproteins.com to learn more and where to buy. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.