{
  "items": [
    {
      "slug": "what-if-singularity-were-a-memory",
      "source_slug": "y-si-la-singularidad-no-fuera-un-evento-sino-un-recuerdo",
      "title": "What if the singularity were not an event, but a memory?",
      "date": "2026-07-30T22:00:00Z",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "AI and business",
      "image": "/images/uploads/techlatte-singularidad-2026-07-30.webp",
      "summary": "A reflection from Piedecuesta on the acceleration of artificial intelligence and the legal, ethical and business challenge of remaining accountable for its decisions.",
      "body": "# ☕ TECHLATTE: IDEAS WITH FOAM\n\n## What if the singularity were not an event, but a memory?\n\nA few days ago, I stopped at a Terpel station in Piedecuesta. While my electric vehicle was charging, I ordered a cappuccino, opened my iPad and began reading the news.\n\nIn his essay *The Gentle Singularity*, Sam Altman wrote that we are already past the event horizon and that takeoff has begun. My immediate reaction was: *what an exaggerated statement*.\n\nFive minutes later, I thought the opposite.\n\nNot because I now believe an artificial intelligence more capable than all of humanity already exists, or that robots are about to rule us. I changed my mind because I realized that we may be imagining the singularity in the wrong way.\n\nFor years, we have waited for it like an eclipse: a precise moment everyone will witness at the same time. Technological revolutions do not work that way.\n\nNobody woke up one morning and said, Today the Internet began. Nor was there a single day when everyone understood that smartphones would transform how we work, shop, travel and relate to one another.\n\nIt simply happened. First, it was a novelty. Then a tool. Later, a necessity. One day, we discovered that we no longer knew how to live without it.\n\nI believe the same thing is happening with artificial intelligence.\n\nWhile we debate whether the singularity will arrive in five, ten or twenty years, AI already writes reports, develops software, analyzes contracts, interprets medical tests, creates marketing campaigns, negotiates with customers and coordinates agents that work with other agents.\n\nThe most surprising part is not that it performs all these tasks. It is that **every month it performs them better than the month before**.\n\nFor centuries, technology advanced because humans invented better tools. Today, tools are beginning to participate in the creation of even better tools. That shift may sound small. In reality, it changes everything.\n\nThe point is not that AI has already surpassed humanity. It is something more uncomfortable: **the speed of technological progress may be surpassing our ability to understand it**.\n\nWhen that happens, the consequences appear not only in laboratories, but in companies, universities, governments and professions such as mine.\n\nAs a lawyer, I see organizations asking every week whether they should adopt artificial intelligence. In my view, that conversation is already behind us.\n\nThe real question is: **who will be accountable for decisions made by an artificial intelligence system?**\n\nAI is no longer merely an experiment. It recommends credit, reviews contracts, detects fraud, screens candidates, drafts corporate policies and participates in decisions that until recently were exclusively human.\n\nThe greatest challenge is therefore no longer only technological. It is legal, ethical and organizational. I suspect many institutions have not yet understood that.\n\nPerhaps twenty years from now, historians will debate the true beginning of the singularity. They may cite a new OpenAI model, an Anthropic innovation or a discovery we cannot yet imagine.\n\nI think they may be looking for the wrong date. Great transformations rarely announce their arrival. We recognize them only when we look back.\n\nAs I finished my coffee, the charger indicated that the battery was full. I put away the iPad and continued my journey, but one thought remained with me.\n\n**Perhaps the day we can identify exactly when the singularity began will be the day we realize we have already been living inside it for years.**\n\nIf that is true, the most important question is no longer when the future will arrive. It is whether we are evolving at the same pace.\n\n☕ **This coffee is served.**\n\nWhat do you think? Did Sam Altman exaggerate to spark debate, or was he simply the first to say aloud what was already happening in front of us?\n\n## Sources and recommended reading\n\n- [Sam Altman — The Gentle Singularity](https://blog.samaltman.com/the-gentle-singularity)\n- [OpenAI — Research](https://openai.com/research/)\n- [Anthropic — Research and AI safety](https://www.anthropic.com/research)\n- Stuart Russell — *Human Compatible*\n- Nick Bostrom — *Superintelligence*\n- Ray Kurzweil — *The Singularity Is Near* and *The Singularity Is Nearer*\n- [Stanford HAI — AI Index Report](https://hai.stanford.edu/ai-index)\n- [World Economic Forum — Artificial intelligence](https://www.weforum.org/topics/artificial-intelligence/)"
    },
    {
      "slug": "chatgpt-called-the-fbi-who-decides-when-to-report-you",
      "source_slug": "chatgpt-llamo-al-fbi-quien-decide-cuando-denunciarte",
      "title": "ChatGPT called the FBI: who decides when to report you?",
      "date": "2026-08-24T12:00:00Z",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "AI and business",
      "image": "/images/uploads/chatgpt-fbi-denuncia-2026-08-24.webp",
      "summary": "The Florida and Tumbler Ridge cases raise difficult questions about privacy, emergency disclosures and who should decide when an AI platform alerts law enforcement.",
      "body": "# ☕ ChatGPT called the FBI: who decides when to report you?\n\n⏱ **Estimated reading time: 5 minutes**\n\nA person opens ChatGPT late at night.\n\nThe conversation begins with a breakup. Then come threats, weapons and, finally, a plan to murder an ex-girlfriend.\n\nAccording to press reports, this happened with Darren Shida Zhou, a 25-year-old man in Florida. OpenAI provided information to the FBI; local authorities opened an investigation, and Zhou received eight years of probation after pleading guilty to charges related to threats against his former partner.\n\nAt first glance, this looks like successful prevention. But the real debate begins afterward.\n\n## The question is not whether OpenAI acted correctly\n\nIt probably did. If a company identifies a concrete, credible and potentially lethal threat, it is difficult to argue that it should simply look the other way.\n\nThe problem is **who decides when a threat crosses the line**.\n\nIn the United States, federal law allows certain providers to disclose communications to authorities when they believe in good faith that an emergency involving danger of death or serious physical injury requires disclosure without delay. This is an emergency disclosure authority, not a specific nationwide rule requiring chatbots to report every threat.\n\nThe law permits action. The company decides when to act. That distinction is enormous.\n\n## Tumbler Ridge shows the opposite risk\n\nMonths before the Zhou case, OpenAI had detected violent conversations involving a user in Canada. His account was blocked. The company considered alerting authorities but concluded that the risk did not meet its threshold for imminence and credibility.\n\nSeven months later, that user killed eight people in Tumbler Ridge, British Columbia. The case is under investigation and has led to litigation. OpenAI later acknowledged that, under updated criteria, the case would probably have been referred to authorities.\n\nTwo users. Two private decisions. Two radically different outcomes. That is the real dilemma.\n\n## The striking legal insight\n\nOpenAI may be becoming something we never imagined: **a private filter positioned before the police**.\n\nAn algorithm detects. A human team assesses. A private company decides. Only then does the State appear.\n\nThe platform does not arrest or convict anyone, but it can activate the machinery that does. That is power.\n\n## Your chats are not protected by attorney-client privilege\n\nThere is also a dangerous misconception.\n\nMany people talk to ChatGPT as they would talk to a lawyer, doctor, therapist or friend. Legally, those relationships are not the same.\n\nA conversation with a commercial chatbot **does not automatically carry professional confidentiality or legal privilege**. Any protection depends on the context, the professional relationship and the applicable rules; it does not arise merely because someone uses an AI tool.\n\nAn AI platform may therefore store confessions, business secrets, family problems, fraud, threats, intimate thoughts or illegal conduct. AI is becoming one of the largest repositories of private disclosures of our time.\n\n## A fragmented regulatory landscape\n\nThe United States does not have one uniform chatbot regime. By June 2026, several states had enacted rules addressing conversational AI, transparency, minors, self-harm and safety. Yet there was still no uniform national standard defining **when an AI system must report threats against third parties**.\n\nThe same conversation may therefore produce different consequences depending on the state, platform, type of threat and internal company policy. For AI companies, this will be a major compliance challenge.\n\n## The risk runs in both directions\n\nIf a platform reports too much, it may undermine privacy, create false positives or involve police unnecessarily. If it reports too little, it may face criticism for failing to warn about a threat.\n\nTumbler Ridge has already opened that judicial and regulatory debate. Platforms may be questioned both **for speaking and for remaining silent**.\n\n## My position\n\nI do not believe the solution is to prohibit AI systems from reporting threats. Nor should private companies have absolute discretion. We need three things:\n\n**1. A clear legal threshold.** What elements turn a disturbing conversation into a genuine threat?\n\n**2. Mandatory human review.** An automated decision should never be enough on its own.\n\n**3. External oversight.** Criteria that can transform a private conversation into a police alert should not exist solely inside a corporate policy.\n\n## If you remember only one idea\n\nWe built machines to which millions of people tell almost everything. Then we installed systems capable of detecting danger. Finally, we allowed a company to decide when a conversation should reach the police.\n\nSometimes that decision may save a life. Other times, it may come too late.\n\nThe right question is not whether ChatGPT should call the FBI. It is: **who should decide when it does, and who watches the decision-maker?**\n\n☕ **This coffee is served.**\n\nWould you accept an AI system analyzing your private conversations if doing so could prevent a murder?\n\n## Sources\n\n- [18 U.S.C. § 2702 — voluntary disclosure of communications or records](https://uscode.house.gov/view.xhtml?edition=prelim&num=0&req=granuleid%3AUSC-prelim-title18-section2702)\n- [AP — OpenAI apologizes over Tumbler Ridge warning failure](https://apnews.com/article/dec2adaad3946583519370eede6a99e2)\n- [AP — lawsuit filed by Tumbler Ridge victims’ families](https://apnews.com/article/26d632a065e8420a263655c8ee29f961)\n- [WPBF — Darren Shida Zhou case in Palm Beach](https://www.wpbf.com/article/south-palm-beach-man-cyberstalking-ex-girlfriend-threats-chatgpt/73434967)\n- [IAPP — state chatbot laws](https://www.iapp.org/news/a/chatbot-laws-coming-to-a-state-near-you)"
    },
    {
      "slug": "cancer-should-not-come-with-a-tutela-action",
      "source_slug": "el-cancer-no-deberia-venir-acompanado-de-una-tutela",
      "title": "Cancer should not come with a tutela action",
      "date": "2026-08-30T12:00:00Z",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "Colombian legal developments",
      "video": "/images/uploads/cancer-tutela-derecho-salud-2026-08-30.mp4",
      "summary": "When a cancer diagnosis arrives, families should be able to focus on the patient—not on learning how to fight the healthcare system.",
      "body": "# ☕ Cancer should not come with a tutela action\n\n⏱ **Estimated reading time: 4 minutes**\n\nSome words change a conversation completely: **It is cancer.** Then come questions about treatment, medication, specialists, prognosis and family decisions. In Colombia, however, many families quickly discover another reality: the appointment never appears, the medicine does not arrive, approval remains pending, the specialist is in another city or surgery is postponed.\n\nSomeone eventually says: You have to wait. The problem is that cancer cannot always wait.\n\n## Cancer patients receive special constitutional protection\n\nIn Colombia, a person with a confirmed or suspected cancer diagnosis benefits from **enhanced constitutional protection**. Since 2024, Law 2360 has expressly recognized this protection. Healthcare should not remain trapped indefinitely in administrative procedures.\n\nFour principles are especially important: **timeliness**, so treatment is received when medically required; **continuity**, to prevent unjustified interruption; **comprehensiveness**, so care is not fragmented into isolated approvals; and **accessibility**, because authorizing treatment is of little value if the patient cannot actually reach it.\n\nAccessibility matters particularly outside major cities. A patient living in San Gil, Barrancabermeja, Piedecuesta, Quibdó, Riohacha, Pasto or another municipality without a required specialized service may also face transport, lodging, caregiver or referral barriers.\n\nThe right to healthcare should not end where a municipality’s installed capacity ends.\n\n## When should a delay cause concern?\n\nNot every administrative difficulty automatically amounts to a rights violation. Families should nevertheless pay attention to medication that is not delivered, surgery without a scheduled date, interrupted treatment, referrals that are never arranged, approvals that do not become actual care, or patients required to travel without the means to do so.\n\nWhen these failures repeat, **the issue may stop being merely administrative and become a legal problem**.\n\n## A common mistake: conversations without evidence\n\nFamilies call repeatedly and explain the case to different officials, but they do not build a documentary record. When they finally need to file a formal claim, they discover they have many conversations and few documents.\n\nA useful rule is: **in healthcare, a phone call may solve the day, but a filing reference can protect the case**.\n\nMedical orders, clinical records, approvals, denials, emails, dates, responses and filing numbers may later become essential to defining a strategy.\n\n## Must a tutela action be filed immediately?\n\nNot necessarily.\n\nEach case requires an initial review of what the treating physician ordered, what is being breached, who is responsible, how much time has elapsed and what risk the delay creates for the patient.\n\nDepending on those answers, different avenues may exist: a formal petition, a complaint before the National Health Superintendence, support from the Ombudsman’s Office or a municipal ombudsman, a specific administrative request or, where fundamental rights are threatened or violated, a **tutela constitutional action**.\n\nFurther questions can arise even after a tutela judgment: Was the order sufficiently clear? Is it being fully complied with? Was only part of the treatment authorized? Should compliance be requested? Are there grounds for contempt proceedings?\n\nThat is where every story becomes different.\n\n## The goal is not merely to complain, but to complain effectively\n\nA family facing cancer already carries enough uncertainty. It should not also have to improvise a legal strategy.\n\nThe central problem is often not proving that the patient needs care—the physician has already established that. The challenge is identifying where the system is blocked and which mechanism may appropriately unblock it.\n\n## If you are facing a similar situation\n\nBefore acting, gather five basic elements:\n\n1. **Relevant clinical records.**\n2. **Orders issued by the treating physician.**\n3. **Existing approvals or denials.**\n4. **Filing numbers and responses received.**\n5. **A simple chronology of events.**\n\nThese documents provide a starting point for understanding the case and evaluating which path may make sense.\n\n## A final reflection\n\nCancer already demands too much from patients and families: time, money, emotional strength and difficult decisions. The healthcare system should not add a second illness: **bureaucracy**.\n\nIf you or a family close to you are facing barriers to cancer treatment in Colombia, we can begin by reviewing the chronology, medical orders and responses to determine which legal avenue may be appropriate.\n\n**Every case is different. That is precisely why the strategy must be different as well.**\n\n> This content is general information. It does not replace medical assessment or individual legal advice. The appropriate route depends on the documents, clinical urgency and circumstances of the case.\n\n## Official sources and case law\n\n- [Statutory Law 1751 of 2015 — fundamental right to health](https://www.funcionpublica.gov.co/eva/gestornormativo/norma.php?i=60733)\n- [Law 2360 of 2024 — special protection for people with suspected or confirmed cancer](https://www.secretariasenado.gov.co/senado/basedoc/ley_2360_2024.html)\n- [Constitutional Court, Judgment T-081 of 2016](https://www.corteconstitucional.gov.co/relatoria/2016/t-081-16.htm)\n- [Constitutional Court, Judgment T-072 of 2026](https://www.corteconstitucional.gov.co/relatoria/2026/T-072-26.htm)"
    },
    {
      "slug": "forming-a-colombian-sas-design-the-company-from-the-start",
      "source_slug": "crear-una-s-a-s-en-colombia-disenar-bien-la-sociedad-desde-el-inicio",
      "title": "Forming a Colombian SAS: design the company properly from the start",
      "date": "2026-09-04T18:00:00Z",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "Corporate governance",
      "image": "/images/uploads/techlatte-sas-colombia-2026/slide-01.webp",
      "summary": "Incorporation can take only a few days. Designing the relationship among shareholders properly can protect a company for years.",
      "body": "Forming a Colombian SAS can look simple: choose a name, complete the forms, register the company and move on. But that is where the real risk begins.\n\n**A company should not be built by copying another company’s bylaws.** Each business has different shareholders, contributions, risks, decision rules and expectations that should eventually be put in writing.\n\nMany shareholder disputes do not arise because the company was registered incorrectly. They arise because nobody decided what should happen if the shareholders disagree, a new investor arrives, someone wants to sell their shares or a major decision divides the owners.\n\nThe flexibility of the SAS is one of its greatest advantages. It also means the bylaws should be treated as **the company’s legal operating system**, not merely a form to complete.\n\nThe accompanying carousel highlights issues to consider before incorporation and why sound design at the outset may avoid more costly problems later.\n\n☕ **Incorporation can take days. A properly designed relationship among shareholders can protect the company for years.**\n\nWere your bylaws actually designed for your business, or merely adapted from a template?\n\n> This is general information and does not replace legal advice for a particular corporate matter.\n\n## Legal source\n\n- [Law 1258 of 2008 — Simplified Stock Companies](https://www.funcionpublica.gov.co/eva/gestornormativo/norma.php?i=34130)"
    },
    {
      "slug": "even-the-people-building-ai-are-afraid-of-it",
      "source_slug": "quienes-construyen-la-ia-tambien-le-temen",
      "title": "Even the people building AI are afraid of it",
      "date": "2026-09-09T12:00:00-05:00",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "TECHLATTE · AI governance",
      "stream": "techlatte",
      "video": "/images/uploads/techlatte-quienes-construyen-ia-2026-09-09.mp4",
      "summary": "Developers of advanced systems also warn that we still do not know how to govern them with sufficient safeguards.",
      "body": "# ☕ TECHLATTE: IDEAS WITH FOAM\n\n## Even the people building AI are afraid of it\n\nA researcher closes his computer and leaves one of the world’s most powerful artificial-intelligence laboratories.\n\nJacob Coxon, a former researcher at OpenAI and Anthropic, says both companies are moving toward systems capable of improving themselves without guarantees that they can be kept under control.\n\nThat warning is serious, but precision matters: his resignation does not prove that AI will destroy humanity. It does show something concrete. People who understand the technology deeply recognize risks they still do not know how to resolve, while competition pushes them to move faster.\n\nAnthropic has safety policies. The governance question is that the same companies developing these systems decide how much risk is acceptable.\n\nAdopting AI requires limits on autonomy, human oversight, traceability and clear responsibility when something goes wrong.\n\nThe question is no longer whether AI will advance, but who controls the race and who will answer if control is lost.\n\n> Would you leave that decision solely to the companies competing to arrive first?\n\n☕ **This coffee is served.**\n\n#ArtificialIntelligence #AIGovernance #TechnologyRisk #TECHLATTE"
    },
    {
      "slug": "private-shareholders-agreements-the-rules-no-one-thinks-about-until-trouble-starts",
      "source_slug": "acuerdo-privado-de-accionistas-las-reglas-que-nadie-piensa-hasta-que-aparece-el-problema",
      "title": "Private shareholders’ agreements: the rules no one thinks about until trouble starts",
      "date": "2026-09-12T16:00:00Z",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "Corporate governance",
      "video": "/images/uploads/techlatte-acuerdo-accionistas-2026-09-12.mp4",
      "summary": "A private shareholders’ agreement can anticipate deadlocks, succession, exits, transfers and decisions that may change a company’s future.",
      "body": "# Private shareholders’ agreements: the rules no one thinks about until trouble starts\n\n⏱ **Estimated reading time: 4 minutes**\n\nTwo people start a company. They trust each other, contribute capital and divide the shares 50/50. They register the bylaws, open a bank account and start selling. For years, everything works—until one dies, or they simply stop agreeing.\n\nWere the bylaws designed for that change? This is where a tool many Colombian companies still underestimate becomes important: the **private shareholders’ agreement**.\n\nArticle 24 of Law 1258 of 2008 allows shareholders of an SAS to agree on share purchases and sales, transfer restrictions, voting, representation and other lawful matters. Its value lies in anticipating issues nobody wants to discuss at the beginning. For the company to be bound under Article 24, the agreement must be deposited at the offices where its management operates. The law also sets a maximum ten-year term, extendable unanimously by the signatories for periods of no more than ten years each.\n\n## Scenario 1: a shareholder dies\n\nImagine two shareholders, each with 50%. One dies unexpectedly. The shares do not disappear; in general they enter the estate and may eventually pass to heirs. A long-standing business partner can be replaced by a spouse, children or other heirs with very different interests. Can the shares be sold? Do the remaining shareholders have a first opportunity to buy? How is the price set? Who represents the shares during probate?\n\nBylaws, shareholders’ agreements, valuation rules, purchase options, insurance and properly designed succession arrangements can address these questions in advance. The goal is not to ignore heirs’ rights, but to avoid turning a family tragedy into a business crisis.\n\n## Scenario 2: a 50/50 deadlock\n\nOne shareholder wants to borrow to expand; the other thinks the risk is too high. One wants dividends; the other wants reinvestment. Neither has a majority. A business with customers, employees and assets may still be paralyzed.\n\nA pre-agreed deadlock mechanism may offer an exit. Under a so-called Russian Roulette clause, one shareholder offers to buy the other’s shares at a stated price, but the recipient may choose to buy the offeror’s shares at that same per-share price. Setting an artificially low price therefore carries a risk for the person who proposed it. Under a Texas Shoot-Out arrangement, both parties may submit bids under agreed rules. These are not statutory one-size-fits-all solutions: their usefulness and enforceability depend on drafting consistent with the law, the bylaws and the shareholders’ circumstances.\n\n## More than two scenarios\n\nWhat if a shareholder wants to sell to a competitor, stops working shortly after formation but keeps a large stake, divorces, has shares attached, becomes incapacitated or uses company information to compete? What if an investor wants the whole business but a small shareholder refuses, or a majority shareholder sells and leaves a minority shareholder with a stranger?\n\nPossible tools include tag-along and drag-along rights, vesting, good-leaver and bad-leaver rules, put and call options, reserved matters, transfer restrictions, pre-emptive rights, valuation mechanisms and family protocols. Each begins with a simple question: what do we want to happen if this situation occurs?\n\n## Why a downloaded template is not enough\n\nCompanies differ in ownership, families, capital, work inside the business, know-how, dividend needs and growth plans. Copying another company’s bylaws or agreement can create the very problem the document was meant to prevent. The bylaws and shareholders’ agreement must work together and reflect this particular company.\n\nDeath, divorce, incapacity, deadlock or a forced sale are uncomfortable topics when everyone is focused on growth. That is precisely why it may be the best time to discuss them.\n\n**If something unexpected happened to any of us tomorrow, would we know exactly what would happen to our shares?**\n\n> This is general information and does not replace legal advice for a particular agreement or corporate situation.\n\n## Legal source\n\n- [Law 1258 of 2008, Article 24 — shareholders’ agreements](https://www.secretariasenado.gov.co/senado/basedoc/ley_1258_2008.html)"
    },
    {
      "slug": "the-real-ai-apocalypse-might-be-jobs-not-war",
      "source_slug": "el-verdadero-apocalipsis-de-la-ia-podria-no-ser-una-guerra-podria-ser-el-empleo",
      "title": "The real AI apocalypse might be jobs, not war",
      "date": "2026-09-16T05:52:41-05:00",
      "author": "Carlos Eduardo Gómez Silva",
      "category": "TECHLATTE · AI governance",
      "stream": "techlatte",
      "video": "/images/uploads/techlatte-ia-empleo-human-reserved-2026-09-16.mp4",
      "summary": "Bill Gates warns about AI’s effect on employment. The challenge reaches beyond job losses to the training of tomorrow’s professionals.",
      "body": "# ☕ TECHLATTE: IDEAS WITH FOAM\n\n⏱ **Estimated reading time: 2.5 minutes**\n\nFor years, we imagined AI danger as something cinematic: machines out of control, autonomous systems, cyberattacks and robots making decisions. Perhaps the more troubling scenario is less spectacular.\n\n**A company discovers that it can operate better, faster and more cheaply with fewer people.** Then another company does the same. And then thousands. This is part of the question Bill Gates has put on the table.\n\n## The problem might not be losing a job\n\nIt might be finding that the job **no longer needs to exist as it once did**. In his essay *The turbulent AI era is here. The choices we make now are critical*, Gates warns that the transition to an AI-driven economy could be especially turbulent because this time automation is reaching cognitive work: analysis, customer service, programming, finance, law, administration and sales. Junior and mid-level workers may feel the effects first.\n\nIf a company automates tasks previously done by a young professional, more than a vacancy disappears. **A step on the learning ladder disappears.** Where will future senior professionals come from if fewer junior roles remain?\n\n## ‘Human Reserved’\n\nGates suggests discussing activities that should be reserved for humans—not because AI cannot perform them, but because it may become very good at them. He mentions care for people and the communication of a serious diagnosis. The same debate could extend to hiring, credit assessment, legal decisions and medical treatment. The question becomes: **do we want AI to do this without a responsible person behind the decision?**\n\n## The economic incentive\n\nPeople require wages, benefits, taxes, holidays, training and time. Software does not operate under the same rules. If automation becomes much cheaper than hiring, businesses will have a powerful incentive to replace human work. Gates therefore raises the possibility of taxing some uses of AI and robots and financing retraining and social protection. The concern is not that technology is the enemy; it is that **economic incentives may accelerate replacement faster than society can adapt**.\n\n## Is a jobs apocalypse inevitable?\n\nWe cannot say that today. The ILO estimates that roughly one in four workers worldwide is in an occupation with some exposure to generative AI, while considering transformation of many jobs more likely than their complete disappearance at present. The World Economic Forum projects both job destruction and creation this decade. But a displaced worker needs to know which jobs will appear, where, when and whether they will have the skills to take them.\n\nColombia has its 2025 CONPES 4144 policy on AI. That is a step forward, but harder questions remain. Which decisions should retain human involvement? Who is responsible when AI harms someone? What happens to workers who lose jobs before they can retrain? How will professionals be formed if entry-level positions disappear?\n\n## The real risk\n\nThe most troubling scenario may not be a machine rebelling. It may be a quiet sequence: one company eliminates ten positions, another a hundred, and young people discover that the roles in which they once began no longer exist. An economy could emerge in which **being able to work no longer guarantees that anyone needs to hire you**.\n\nWe should ask not only what work AI can do, but **which work and decisions we are willing to protect before it is too late**.\n\n☕ This coffee is served. If we had to designate three activities as ‘Human Reserved’ tomorrow, which would you protect?\n\n## Sources\n\n- [Bill Gates — The turbulent AI era is here](https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make)\n- [ILO — Generative AI and jobs: A 2025 update](https://www.ilo.org/publications/generative-ai-and-jobs-2025-update)\n- [World Economic Forum — Future of Jobs Report 2025](https://www.weforum.org/publications/the-future-of-jobs-report-2025/)\n- [DNP — CONPES 4144 of 2025, National AI Policy](https://colaboracion.dnp.gov.co/CDT/Conpes/Econ%C3%B3micos/4144.pdf)"
    }
  ]
}