RECORDNEWS — broadcast
20261004 033000 UTC
513 transcript segments
Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking
Data courtesy of The GDELT Project (https://www.gdeltproject.org/), from the Internet Archive TV News Archive.
Machine transcription. Treat it as a searchable index of what was broadcast, not a verbatim quotation record.

[00:00:00] accounts for nothing. R$5000 is obvious, it stinks bad. The bank doesn't block it,
[00:00:03] and who owes the guy the money? So these
[00:00:07] applications made it easy, you can overturn this. So there is, but you have to
[00:00:12] Then suddenly the file a lawsuit, that's why I say they have to. Because if he called the
[00:00:15] bank and said it wasn't me, and the bank looked and said, let's see the IP. Oh, it really is
[00:00:19] the IP from Rio de Janeiro. The user is from
[00:00:24] content that immediately generates Curitiba. Many strange things. The bank made a mistake. Oh, let's pay, we'll return the amount or not
[00:00:27] charge you. Solution. You have to file a
[00:00:30] lawsuit in 90% of cases. And
[00:00:33] what changes today in the relationship
[00:00:37] between offices, especially
[00:00:41] due to the advancement, especially
[00:00:45] of AIs, right?
[00:00:47] Because today you take a contract, you put it in
[00:00:51] GPT, it gives you the main
[00:00:53] information, it helps you
[00:00:57] write contracts. What changes?
[00:01:00] What changes in offices? What changes in the legal process
[00:01:03] and is it taking away jobs? Not yet, not yet. I
[00:01:07] think it will take away part of the, it will replace labor. I
[00:01:11] am very, I
[00:01:13] and her husband's. The husband started I strongly believe in this thesis, in fact I have a lecture that
[00:01:16] worries me. I think that labor will partly be, who
[00:01:20] will lose? A large part of human
[00:01:24] functions will be replaced by generative AI.
[00:01:27] to unlock it. It's like he took a locked phone, and So many people will lose, especially... those who do not
[00:01:30] depend on, for example, right away,
[00:01:34] voice actors, screenwriters,
[00:01:36] uh, who, who makes, you're going to make, you're going to
[00:01:40] make people here start crying. No, but it's, videomakers, we
[00:01:44] are a TV, there are, there are all these people you're talking about, uh, but did you see
[00:01:48] that Disney screenwriters were on strike, Hollywood screenwriters were on
[00:01:51] strike, because they know that many movies are being written by AI
[00:01:55] that is good, book authors. So for the people here
[00:01:59] in the production not to worry too much or not to like me, I have another
[00:02:02] thesis: today the story of handmade is more expensive, you want a
[00:02:06] handmade shoe, it's the luxury of luxury, I think the future will be that
[00:02:10] handmade will be the luxury, so ah there's the script made by AI and there's the one
[00:02:16] and you know that here you can trust in that will be by the screenwriter who is a man who is a human being, I say not that it's a person, ah so let's do it with a
[00:02:20] person, he has more human
[00:02:23] sense, right? More sense, so I think maybe the future will be this handmade story,
[00:02:27] but regarding offices, do we have time? for me to tell you or will it be
[00:02:31] next? We have time. Regarding offices, uh, for
[00:02:34] now it is not, it is not replacing
[00:02:38] it in law and it is hindering, hindering,
[00:02:41] why? Because the client relies on the AI's
[00:02:45] answer as if it were a legal consultation. The AI makes
[00:02:49] many mistakes for Brazilian law. The
[00:02:52] ChatGPT has indexed content in Portuguese close
[00:02:56] to 1 to 5% only. ChatGPT is a platform
[00:03:00] that indexes content in English 95%, so it is much
[00:03:03] better for issues related to
[00:03:06] literature, American law,
[00:03:10] English law, in English. So much so that for medicine it is much
[00:03:13] better than for Brazilian law.
[00:03:17] In our, in our area it makes many mistakes
[00:03:19] and when it makes mistakes, who is to blame? The office
[00:03:23] no, when it makes a mistake, the problem is the AI's problem, because it doesn't
[00:03:27] have, it doesn't have responsibility for the content. So much so that
[00:03:31] if you petition something made by an AI, the responsibility is yours. And
[00:03:35] why does it hinder? Because the client comes with a ready
[00:03:38] answer and when you contest the AI, he starts to doubt your
[00:03:42] opinion. Then the legal opinion becomes
[00:03:45] just a confirmation that the AI was right, but when the AI is
[00:03:49] wrong he has difficulty swallowing that he is wrong
[00:03:53] the platforms and placing the because he wanted it. It's the story of the confirmation bias of fake news, he
[00:03:56] is doing that with AI, including he manipulates the AI with prompts.
[00:04:00] Because let's say he's thinking about separating from his wife then he's there
[00:04:04] asking and he manipulates it because he asks several questions and then she answers, and by the
[00:04:08] end she answers more or less what he wanted, which often ChatGPT and
[00:04:11] AIs do that, they give you what you want even if it's wrong. Then he goes
[00:04:15] for legal consultation, the lawyer says you have no right, he goes crazy, saying how can I not
[00:04:19] have it, I asked her and she said I have the right. And then you know what he
[00:04:23] will do? Look for another one until someone confirms his thesis. But sometimes
[00:04:27] none will confirm. So today it is hindering. I received many
[00:04:30] emails because I disclose my private email on
[00:04:35] a company, except for environmental crimes, which is websites and so on. I receive emails that clearly look like ChatGPT made them. The client already
[00:04:38] sends them ready, the script is badly made because the AI does not
[00:04:42] work well with chat with Brazilian law. And another thing, it's not
[00:04:45] a market reserve, do a test, take
[00:04:49] 30 lines of values, I did this with, for example, a credit card
[00:04:53] statement and ask it to
[00:04:57] where does add them up. For me, every time I did this, ChatGPT made a mistake. It gave the wrong
[00:05:00] calculation. Then I say, my account showed so much, she said, ah, I really checked here,
[00:05:04] you're right. The other day I asked a question
[00:05:08] about a client we wanted to do a survey, and so on. He said
[00:05:11] that his mother... he gave the name of the client's mother as his sister, he made a mistake.
[00:05:15] Then I say, but this person is not his mother. Ah, I saw here that I really looked
[00:05:18] better. For me, ChatGPT or AI today is the lazy
[00:05:22] intern, does everything carelessly, makes too many mistakes, you can't
[00:05:25] trust it. But people give it much more credibility than
[00:05:29] it should have. You mentioned LGPD, right? It
[00:05:32] changed the way people protect data, people,
[00:05:36] but there are companies that still treat it in a
[00:05:40] bureaucratic way. How
[00:05:43] come, because every
[00:05:45] now and then
[00:05:48] there are data breaches, right? Many,
[00:05:51] many from companies that
[00:05:54] have a certain protection, invest in protection, but there are many companies
[00:05:58] that we end up registering that you see that it leaked
[00:06:02] and the guy didn't care about your data. How do you fight
[00:06:06] that? Just to make a caveat for the people, in my opinion,
[00:06:09] it's not because I defined that AI is a
[00:06:12] lazy intern that it doesn't do some things very well. So, for
[00:06:16] example, as an accessory, it's excellent, I use it a lot for my car, it broke down, I
[00:06:20] send it to ChatGPT there, help me. So there is a good side, okay? It's important to
[00:06:23] record that. Speaking about LGPD, in the end, the
[00:06:26] fault was that the law was good, the law brings
[00:06:30] an interesting legal basis, brought legal certainty, it's good for everyone,
[00:06:34] it's like a consumer protection code, even companies liked it to have legal
[00:06:37] certainty, but as the national authority today, the national
[00:06:41] data protection agency has never
[00:06:45] properly supervised, so much so that it has applied three fines in its history, there are 5 years of LGPD, there are
[00:06:49] three fines in its history. You still see leaks almost every day, people didn't pay much
[00:06:52] attention, especially small and medium entrepreneurs. Large
[00:06:56] entrepreneurs are very much charged, right? They end up, another
[00:07:00] thing, large entrepreneurs realized that
[00:07:03] investing in data protection was marketing, not
[00:07:07] an expense. So, for example, I have a story, I
[00:07:10] arrived, I went to New York, I was in Times Square, I was looking at the
[00:07:13] advertisements, three consecutive advertisements were about data
[00:07:17] protection. One was from Meta saying ah, in our,
[00:07:20] in your WhatsApp conversation, it's secure here and so on. The
[00:07:24] other was from Apple. Apple talked so much about data security that
[00:07:28] they transformed an apple into a padlock, I don't know if you saw the advertisement where it turns into a
[00:07:31] padlock, a little apple. Why? My phone is secure, my
[00:07:35] phone doesn't let your data leak, data protection is today a marketing
[00:07:39] topic more than consumer protection. Of the data
[00:07:43] subject. Those companies that saw this surfed a little
[00:07:46] on this wave, invested as uh a
[00:07:50] situation in which it brings security, but it sells this, it has a return, it's
[00:07:53] worth it. But there are many people who don't see it, they say this is bureaucratic, this is an expense
[00:07:57] and so on, screw the consumer, until there's a leak and they lose the consumer, right? But we're
[00:08:01] talking about billion-dollar companies, right? But there are small offices,
[00:08:04] there are small companies that often don't have
[00:08:08] money to invest in security, right?
[00:08:11] Uh and sometimes... you don't even need much money,
[00:08:13] right? What is the guidance for
[00:08:17] these, for these small businesses to reduce the risk of
[00:08:21] precisely data breaches in cyber
[00:08:25] attacks? Micro small business, depending on the size, it
[00:08:28] even has the exemption from LGPD, it doesn't even need to
[00:08:31] comply, because for example, LGPD requires you to have
[00:08:34] a DPO, right? Which is the data protection officer. You,
[00:08:38] you are a bakery, you have a computer that records bread sales, you don't need to have a
[00:08:41] DPO. So there are several uh several
[00:08:45] exceptions, there are layers, there are layers. Now, if you are already a
[00:08:48] company that needs to adapt, I would say
[00:08:51] it's worth thinking about applying LGPD in an
[00:08:55] initial and conservative way, without, of course, spending a truckload of money, because it's
[00:08:59] good for everyone, it's good, but you need a specific person for that. Ah,
[00:09:02] do you need to? Yes, you do, because equipment is... imagine this: if
[00:09:06] you were a lawyer, if I were to set up a small company,
[00:09:09] what kind of equipment would I need? It depends. It
[00:09:12] depends, it depends on the amount of data you're going to process. If
[00:09:16] you're a small credit card company, you need a huge
[00:09:20] backup, huge. Imagine that every credit card operator
[00:09:24] can never let information get lost, otherwise you will
[00:09:28] lose the information of the sale, of the installments of a purchase
[00:09:31] and so on. So it has to be inviolable, almost, right? It has to be very secure. That's why the Brazilian
[00:09:35] banking system is a model for the world. But if you're a small company, you're
[00:09:38] just starting out, you have to have at least an IT person. The
[00:09:42] guy says, "Look, install antivirus, don't leave a computer so vulnerable,
[00:09:46] let's make access layers." We've already had problems, big company. Look, I'll
[00:09:50] give you an example, we had
[00:09:52] uh during the
[00:09:55] pandemic a very serious leak of a
[00:09:58] prescription of
[00:10:02] medication from Dr. David Wiip. He
[00:10:05] was head of the
[00:10:09] combat team, I worked with him, he's a friend.
[00:10:12] A great friend, he was head of the combat team
[00:10:16] there at João's. That's where the story came from, he was Secretary of
[00:10:20] State, Secretary of Health and coordinated the group
[00:10:23] of doctors. He, Jean
[00:10:25] Gorins.
[00:10:28] Then they said, look, he prescribed chloroquine or ivermectin, one
[00:10:32] of the medicines being debated at the time, I think it was chloroquine, chloroquine.
[00:10:37] prescription appeared publicly, this prescription signed by him and so on leaked from
[00:10:40] a large pharmaceutical company, right? From a large
[00:10:43] pharmacy, uh, a compounding pharmacy, no,
[00:10:46] not a chain pharmacy, ah, it wasn't
[00:10:49] a compounding pharmacy, no, no. So much so that we handled this case, we opened an
[00:10:53] inquiry, we found out who leaked it, who it was for, the manager of the
[00:10:56] pharmacy. Damn, the guy who
[00:10:59] puts this information in the system cannot leak it, so much so
[00:11:03] that ah, many people said it's good that it leaked, but that's a crime, he answered for the crime,
[00:11:07] the manager, it can't leak, that's, that's
[00:11:10] data protection, because we think of data protection as name, ID, no,
[00:11:14] sometimes it's an exam you're taking, sometimes it's an exam of a
[00:11:17] pregnant mother who doesn't want to tell her husband to make a
[00:11:21] surprise, the father receives an email offering diapers already,
[00:11:25] there have been data breach processes because the mother
[00:11:28] started buying pregnancy tests,
[00:11:32] she put it on her CPF, right? For the cancer, then the CPFs were linked, hers
[00:11:38] receiving a lot of advertisements for diapers, pacifiers,
[00:11:41] bottles and so on, to the point that he started to suspect that something was wrong and
[00:11:44] later he found out that he would be a father, and this affects reputation,
[00:11:48] the survival of a company, is there any example of a
[00:11:52] leak that somehow harmed the company's image
[00:11:55] and it lost uh lost revenue? But
[00:11:59] we're going to a short break and you'll answer:
[00:12:25] "Trust comes from dialogue, from the search for
[00:12:29] truth, from data journalism,
[00:12:32] exclusive data obtained via the Access to Information Law by
[00:12:36] Record News, from investigation,
[00:12:41] agility, in updated
[00:12:43] information, with context and analysis. If war is
[00:12:46] declared, if laws change, if the mosquito kills, if a sting
[00:12:50] saves, we show it. If the economy advances, if Congress
[00:12:54] splits, the forest burns, if artificial
[00:12:57] intelligence threatens or helps, we discuss
[00:13:00] and you participate. Send us your suggestions.
[00:13:03] Ecor News, you
[00:13:07] trust."
[00:13:15] Returning, Durso, in the previous block, we were talking about
[00:13:18] data breaches and treating them legally, but they
[00:13:22] can also somehow affect the image of
[00:13:25] companies. Do you know, does it affect the image?
[00:13:29] I think in Brazil, uh,
[00:13:32] it affects, but not to the point of bankrupting the company. I've
[00:13:36] never, I've never seen it happen here, perhaps because we are so
[00:13:39] used to leaks and impunity that people
[00:13:42] have kind of swallowed it already, but for example, in the United States
[00:13:46] there is research that when there is a large data breach at a
[00:13:49] company, 40% of consumers do not re-enter
[00:13:52] their data or do business with that company. So a very big blow,
[00:13:56] imagine losing 40% of your consumers, almost
[00:13:59] half. Here in Brazil I am sure that this number does not reach this
[00:14:03] level, but I think people are bothered, people sometimes even
[00:14:07] think about suing, so much so that this topic reached the STJ
[00:14:11] in the end, the STJ's decision was favorable to companies, because just
[00:14:14] leaking data does not immediately generate an obligation to pay
[00:14:18] compensation. So you received a message from the phone operator saying:
[00:14:22] your data leaked through here, if you file a lawsuit you
[00:14:25] don't win the case. Now, if because of this leak
[00:14:29] it fell into the hands of a criminal and a criminal committed fraud, then the company has
[00:14:33] to pay. The problem is proving that your leak generated the fraud, in this
[00:14:36] causal link, which is complex. So it's not that
[00:14:40] easy, so this discouraged many lawsuits. So today in Brazil, in the
[00:14:43] end, because of an ineffective NPD,
[00:14:46] we have a weak LGPD. The law as a whole is, as
[00:14:50] a theory, great, but in practice it's weak, and companies that are not
[00:14:53] very concerned. Was it worth applying the law? Yes, because many companies
[00:14:57] adapted, especially big techs, especially
[00:15:00] billion-dollar companies, as we said, but
[00:15:04] small and medium ones are not very concerned. Where
[00:15:07] does, where does freedom begin? Because
[00:15:10] often they always treat information
[00:15:14] when it is canceled as censorship.
[00:15:17] Where does freedom of expression begin and
[00:15:22] legal responsibility begin? That's a good question, isn't it? Today
[00:15:25] today people, because people say silly things, somehow
[00:15:29] confuse people and then they go
[00:15:32] and get legally burned. Ah, I'm being censored, that's the
[00:15:36] justification today. It's "I'm being censored," right?
[00:15:39] But it's something that often confuses people, leads
[00:15:43] people to make mistakes. The guy says it's not censorship, he has
[00:15:47] the right, it's legal, that's a good question, because in Brazil we
[00:15:51] have to enter this debate looking at two scenarios, the scenario of an abuse
[00:15:54] of the judiciary, especially the supreme court, and a scenario of
[00:15:58] user abuse. So on the one hand, we often have a supreme court
[00:16:01] that silenced people who shouldn't have been silenced, an example is Monark. I have
[00:16:05] all my reservations about talking about Monark. But the supreme court ruling that he cannot
[00:16:09] create any channel ever again in his life, if Monark
[00:16:13] wants to have a channel today to comment on fashion shows, he cannot
[00:16:17] have it due to a decision by the supreme court, by Minister Alexandre de Moraes, this is censorship,
[00:16:20] prior censorship, why? Because if he said something silly
[00:16:24] there on the podcast about Nazism and so on, he should
[00:16:28] be punished, he could even have a sanction of not
[00:16:31] being able to have a podcast anymore. Now, if he wants to have a gaming channel, a fashion
[00:16:35] channel, a political commentary channel,
[00:16:37] whatever it is, he cannot, this is censorship and people are reading this, they say,
[00:16:41] gosh, there's abuse here. On the other hand, people think they can
[00:16:45] say whatever they want, they cannot have the content excluded,
[00:16:48] which is also a mistake. Where is the limit of freedom of expression? Does freedom of expression
[00:16:52] have limits? Yes, the limit is in the
[00:16:55] law itself. The limit of freedom of expression is provided in
[00:16:59] law, which is from the moment you use this constitutional
[00:17:02] guarantee to commit a crime, you are exceeding
[00:17:06] the limit, so ah... "I'm being prejudiced,
[00:17:10] but it's my freedom of expression," no it's not, so much so that by law
[00:17:13] you cannot be prejudiced, so you are committing a crime, you abused your freedom of
[00:17:17] expression and will answer for it. What is the limit? The limit is the right of others, which
[00:17:21] is provided in law in crimes, especially in crimes.
[00:17:25] From the moment the user made a publication, it is criminal, it needs
[00:17:28] to be removed, it needs to be either by the platform,
[00:17:32] through the algorithm or by a judicial decision. If a judicial
[00:17:35] decision comes to remove that publication, it must be
[00:17:39] removed, yes, and it is right to be removed. Is he being
[00:17:43] censored? No,
[00:17:45] no, the judge has this power, this prerogative to, upon seeing a crime,
[00:17:49] eliminate it, prevent it, and that is not
[00:17:53] censorship, it is merely the application of what is provided in law. We talked a lot
[00:17:57] about deepfakes and AI in
[00:18:00] content production in electoral campaigns and in
[00:18:04] changes in electoral campaigns, but
[00:18:07] the world is not just electoral campaigns, we are in an election
[00:18:10] moment, is there a legal
[00:18:14] instrument to combat this? Because, right, we often
[00:18:17] see artists being
[00:18:21] right, uh, somehow violated
[00:18:24] by artificial intelligence videos, is the legal
[00:18:28] world prepared for this? Is there a tool that we talked about elections, but that's
[00:18:32] fine, but the world, the
[00:18:35] world in general, the legal world is prepared to
[00:18:38] defend? In Brazil, no, so much so that no law related to
[00:18:42] AI has been approved, we have nothing related to
[00:18:46] AI approved, we have a law, a bill for
[00:18:49] artificial intelligence that did not pass in congress,
[00:18:53] it's stalled. So in Brazil no, but I see
[00:18:56] ways to combat it, right? It's possible to combat it, yes. I think three
[00:19:00] pillars of combat. First pillar against the user
[00:19:03] author, he must answer for the crime committed with the use of AI.
[00:19:08] Second pillar, through blocking on
[00:19:11] AIs,
[00:19:13] so for example, the individual registers on chat GPT, sends a photo of himself and says, "Make a
[00:19:17] video of Hilsinho asking for votes for such a candidate," it has to say no, I can't, I won't
[00:19:20] produce it. If it doesn't produce it, it's harder for the user to
[00:19:23] viralize. So against the user, against
[00:19:27] uh by placing blocks directly on
[00:19:31] responsibility on social media where the content takes shape
[00:19:34] and is found by other users. If the content is
[00:19:38] illicit, there was a complaint and it is clearly illegal, you can see that it
[00:19:42] is an AI or whatever, it should be removed with more celerity, it should
[00:19:45] be taken down faster. But are you one of those who think that social
[00:19:49] media should be punished in some way for what
[00:19:53] users publish? I am one of those people, but I am not one of those
[00:19:57] who applauded what the supreme court did, because what did the supreme court do? When the supreme court
[00:20:00] ruled on platform responsibility, it stated that the platform today must
[00:20:04] pay for any illicit act, it only removed crimes against honor
[00:20:08] but everything else, I even said many times in an interview
[00:20:11] there with Emílio on Pânico and so on, I said the supreme court
[00:20:15] abused, went too far, why? Because
[00:20:18] do you know where the supreme court should have protected itself? For content
[00:20:22] that is manifestly legal. What is manifestly legal? That
[00:20:25] any human being, upon seeing that content, would say: this should never be
[00:20:29] online, for example,
[00:20:31] child pornography, uh, violence against
[00:20:35] animals, someone killing a dog. This content is criminal, it's
[00:20:39] obvious, it has to come down, incitement to suicide, to terrorism,
[00:20:42] etc. You look at it, you see that it's criminal,
[00:20:45] it has to come down. This should generate responsibility for the platform if it receives a
[00:20:49] complaint. What do I think the supreme court got wrong that went
[00:20:53] further? It included all crimes, then there's
[00:20:56] assault on the democratic state of law. What is assault on the state? Anything?
[00:21:00] That's subjective. I think subjectivity is
[00:21:04] dangerous, so for me, the subjective should have stayed online and become
[00:21:07] a judicial process so that someone could then decide whether to remove it or not, and not as the supreme court
[00:21:11] did, to put it as if it were
[00:21:14] responsibility. And because this, just to conclude, why was I against including
[00:21:18] everything? Because subjectivity creates the risk of censorship. Because I see your
[00:21:21] content there speaking ill of me, I say, "Ah, I didn't like what Som
[00:21:25] said, I'm going to say he's persecuting me," there's a crime of persecution. We must moderate so it doesn't
[00:21:28] become censorship. In my thesis, it's very easy: if it's
[00:21:32] manifestly legal, the platform has to moderate; if it's subjective,
[00:21:36] it goes to the judiciary. So, for example, you spoke ill of me, it's
[00:21:39] subjective, right? I said the crime of persecution, is it or isn't it? It goes to the
[00:21:43] judiciary, the platform doesn't have to remove it. Now, you come and say, "Ah, I
[00:21:46] think Durso should be killed," it's obvious that you want that, this should have... if
[00:21:50] the platform should not be held responsible, that's my thesis. Is it in force?
[00:21:54] No, what the supreme court put is all illicit acts with the exception
[00:21:58] of crimes against honor and this will generate censorship, I also spoke of the elderly
[00:22:02] too, but I
[00:22:04] wanted so that it doesn't seem like only the elderly,
[00:22:07] how can people in general
[00:22:10] protect themselves
[00:22:13] from digital crimes, right? We talked about the elderly, but let's
[00:22:16] go, people in general, because many... it's what you
[00:22:20] said, the voice is very similar, right?
[00:22:24] And I'll give you an example, my mother, about
[00:22:28] it must have been about good
[00:22:31] it wasn't digital, it was by phone, about 25 years ago,
[00:22:35] no, they called her, "Look, auntie," and she
[00:22:39] said the person's name, Eduardo, then the guy said, "Yes, it's me,
[00:22:42] auntie, look, I crashed the car, I need money,"
[00:22:46] and there was no Pix, nothing, the guy went to my house, my mother...
[00:22:54] to get the money. Then I found out because one day I was doing a postgraduate course at
[00:22:57] GV in environmental
[00:23:00] law, I was a sub-prefect more than 20 years ago
[00:23:03] and then a friend called me, "Hey, your mom's calling
[00:23:07] me asking for money here, but I
[00:23:09] didn't lend her money, she's saying, 'Oh, don't do this to me,' "
[00:23:13] and so on. I called my mom, "Mom," then she, she started telling me and
[00:23:16] I said, "Of course, you called the name..." of the person, right?
[00:23:20] "You are a lady, everyone calls you 'auntie', right? You fell
[00:23:24] for a scam. And how do you deal with it in the digital world, right? Man,
[00:23:28] it wasn't even digital. How do you do it for the general public,
[00:23:31] let's leave out the elderly so as not to seem prejudiced, to
[00:23:35] defend themselves, because that's it, today
[00:23:37] AI imitates the voice of the person,
[00:23:41] right? Brad Pitt is dating
[00:23:44] a 70-year-old Brazilian lady here, right? But it's his
[00:23:48] face, it's his video. it's his, right? It's
[00:23:52] fine that an older person comes back, but how do people protect
[00:23:55] themselves? Oh, look, there's a
[00:23:59] suggestion for scams in general involving financial values
[00:24:03] when someone pretends to be a known person that works very well and
[00:24:06] will probably work for a long time, which is like this:
[00:24:09] you get home, you call your family, let's say, let's agree on a
[00:24:13] keyword, so let's say we move a lot of money among
[00:24:17] ourselves, Wilson said, we're going to agree with... our keyword is
[00:24:21] Corinthians, then someone pretends to be you,
[00:24:24] "Oh, I need money, blah blah blah", you don't know "Tio Ká", what's the keyword?
[00:24:28] The criminal will never know, never, even using your voice, even using your
[00:24:32] image, even making a video call, so
[00:24:35] for family members, the keyword is very efficient because it will
[00:24:38] inhibit all use of AI, and since it was
[00:24:42] agreed upon in person, there's no way for it to leak, so the
[00:24:46] use of a keyword is quite interesting, of course it's not
[00:24:49] perfect because the criminal can pretend to be the mother
[00:24:53] first, call the family member, check the keyword, then apply the
[00:24:57] scam, so it's not immune, you have to be careful before giving away the keyword,
[00:25:01] but it still works very well, you know, something I do on my
[00:25:04] phone that I stopped doing, I moved, my mother's name
[00:25:08] is her name, it doesn't say mother, oh that's right, that's also good, I won't
[00:25:12] say the name because then someone might want to get into my phone and know my mother's name, but
[00:25:16] I put my mother's name, I put my brothers' names, father-in-law... I put
[00:25:20] my wife, I don't put wife, I put nothing because if the guy invades, it's
[00:25:24] easy to know it's a son, then he takes the phone and sends it to
[00:25:28] the guy who fell for it,
[00:25:30] he fell for it because it also said "dad" on your
[00:25:33] son's phone, right? It said "dad", so then the guy
[00:25:37] came in and said "dad, lend me money, there's a bill
[00:25:40] here," you know, that also
[00:25:44] helps, and that's what helps, right? Everything that
[00:25:48] makes the criminal's life bureaucratic helps. I'll tell you what I do with my phone, look how crazy
[00:25:51] it is, the staff, I think if they want and are interested, even on my Instagram, Luiz
[00:25:55] Augusto Durso, I made a video teaching how to do this. On my phone,
[00:25:59] like this, every time someone tries to put
[00:26:03] my phone in airplane mode, which is an operating mode when someone steals
[00:26:07] or robs a cell phone, they immediately put it in airplane mode. If they take it unlocked, they
[00:26:10] put it in airplane mode, and if they ask for your password in a robbery, they put it in airplane
[00:26:14] mode. My phone blocks the screen and reactivates the line.
[00:26:18] It never enters airplane mode. It's a shortcut I made for iPhone, and it
[00:26:21] will never enter airplane mode. Second thing, if I send a text message
[00:26:25] from any number to my phone saying "block," it will block
[00:26:29] the screen because I made a shortcut that every time it receives a message saying
[00:26:32] "block," it blocks the screen. So let's say I'm here, where do you send the message? In
[00:26:36] a text message. If I take your phone and send a text message to my phone saying "block," it will
[00:26:40] block my phone screen, just that day of putting the password on the screen.
[00:26:44] But how do you do that when you're rich, when you have an iPhone? I have
[00:26:47] a Samsung. No, Android also has these shortcuts. They are
[00:26:51] applications that you configure for some kind of action, logically some
[00:26:55] will have other options and so on, but it's worth researching. At
[00:26:59] least for iPhone, you can do it, I did it. So
[00:27:02] why? What is this for? Imagine
[00:27:06] I'm here at the studio door
[00:27:10] requesting a ride-sharing app, a kid passes by on a
[00:27:12] bicycle, my phone is unlocked, he doesn't know my
[00:27:15] password, he took the phone. If he puts
[00:27:19] it in airplane mode, the screen will lock. So if he puts it in airplane mode, it's
[00:27:22] already solved. If he doesn't put it, I'll send a text message
[00:27:26] and it will lock the screen, and he doesn't have a password
[00:27:30] taking a locked phone, then it's a problem? It locked the screen
[00:27:33] like that, right? If you get it back, you put your screen password and you
[00:27:37] won't lose anything, you won't lose anything unless you go to Find My Android -
[00:27:41] Find My iPhone, then you erase the data, which is also important. Many people call
[00:27:44] me, this is important, people call and say, "Durso, my phone was stolen
[00:27:48] or pilfered," and I say, "Okay, go to Find My iPhone
[00:27:52] or Find My Android and block the phone, then erase the data." They say, "Ah, but I
[00:27:56] never configured that." This has to be configured, they must have
[00:27:59] configured it but don't even remember the password, that's a problem. So I'm giving you a friendly
[00:28:03] suggestion here: when you get home, do a password restore
[00:28:06] and write it down at home, why? Because if your phone is stolen, you'll be able
[00:28:10] to erase the data manually from your computer. This is for any
[00:28:14] iPad, any phone you log into, you go to Find My Android if it's
[00:28:17] Android, it's it's
[00:28:19] android.com/find. For iPhone, it's www
[00:28:23] icloud.com, you'll log in with your username and password and you'll
[00:28:27] be able to locate the device, so it's good because you can know where it is and you
[00:28:31] can block it if he puts it in airplane mode, you create shortcuts to
[00:28:34] block it and so on. So I did that, you pull it up, it's crazy, yeah, but
[00:28:38] I have it on my screen, that's cool, I'll show you when we finish. In the
[00:28:42] phone apps, there's a bank app. If you click
[00:28:46] on it, it locks the phone, it's fake, it's just to lock the phone, because if the
[00:28:50] guy takes it and somehow has signal, he'll click to open the bank and it will
[00:28:54] lock. So I create these things, I keep researching, I use my Instagram
[00:28:57] to share this information, so everything
[00:29:01] you make bureaucratic for the criminal's life is better. What
[00:29:05] is the biggest challenge the legal world will face with
[00:29:09] technology and cybercrimes in the next 10 years?
[00:29:11] It could be crime or artificial intelligence.
[00:29:15] Artificial intelligence, by far. Why? Because it's not just what we're
[00:29:18] seeing about the use of AI, but
[00:29:20] AI committing crimes. So, for example, soon autonomous cars will be
[00:29:24] running over people, and an autonomous car can
[00:29:28] have milli-microseconds and milliseconds to decide.
[00:29:32] Then someone will have to program a car so that if it's
[00:29:36] on a road and has to kill an old lady or a baby, it will have to
[00:29:39] decide. Humans act on impulse. Impulse
[00:29:42] is not usually criminalized because we
[00:29:46] understand that this is natural for human beings. Then, "Oh no," then
[00:29:49] it has a much faster decision response
[00:29:52] and it should run over. I ask you, this will be a legal
[00:29:56] dilemma: this autonomous car that has no driver, should it run over the old lady
[00:30:00] or the baby?
[00:30:03] There's no right answer. Right, there isn't, but someone will have to
[00:30:06] program it. So, these AIs, tomorrow AI will hallucinate
[00:30:10] and it will be working in a company, it will throw an employee against the wall
[00:30:13] and commit bodily injury, which has already happened, right? Today, there was an AI
[00:30:17] that was playing chess with a boy and broke the boy's finger. It hallucinated, broke
[00:30:21] the boy's finger playing chess, bodily injury
[00:30:24] committed by an AI. It's not that someone programmed it to do that, it's not that it's an
[00:30:27] instrument for crime, it automatically committed a crime because it wanted
[00:30:31] to. And then how does it respond? Who will respond? The company, because there's no
[00:30:34] criminal liability for
[00:30:38] the criminal liability of the legal entity. So it's not just the use of AI for
[00:30:41] crime, it's the use of AI committing
[00:30:44] crime. Also, there's a general challenge out there, which is
[00:30:48] quantum computing with regard to cryptography. I
[00:30:52] think in 10 years we will have evolved a...