Metta has discussed cutting off all funding for its independent oversight board by the end of their current commitment. That is pulling the plug on a corporate governance experiment designed specifically to take the heat off Mark Zuckerberg. Yeah, I mean that board was built to handle high-stakes speech decisions. They were basically the Supreme Court for the platform, but now Metta is aggressively restructuring. They are cutting 15,000 jobs and pivoting entirely toward artificial general intelligence. So if a global platform dismantles its human safety net to fund automated intelligence, how does it govern the flow of information for billions of users? Well the engine driving this internal restructuring is a really rigid performance review system. Managers are actually forced to place 15 to 20 percent of their staff into a meat's most or lower category. Wait, hold on, back up. Yeah. Are you saying that even if a team consists entirely of high performers, a manager is forced to penalize a set percentage of them just to meet a quota? Exactly. The system forces a distribution curve. Historically, the performance cycle was a mechanism for growth, you know, getting feedback, calculating bonuses. Right. But during workforce reductions, it becomes the primary mechanism for contraction. It is like grading on a curve in a class of valedictorians. Even if everyone scores a 99%, someone still has to fail just to satisfy the math. It basically turns teammates into competitors overnight. And that completely destroys the concept of a meritocracy. The internal culture changes into a survival game. An employee who gets that meat's most rating faces a severe bonus cut and they land on a stealth list for the next round of layoffs. I mean, the phrase, meat's most, sounds perfectly fine in a normal corporate environment. It implies you are doing your job. Right. Usually it does. But in this specific mathematical curve, it is a death sentence for your career. If I am an engineer working there, I am terrified. You are essentially sacrificing your own crew just to maintain momentum. You really are. So you adapt your behavior to stay on board. Surviving crew members are abandoning legacy projects entirely. Whole teams are fleeing the virtual reality division, reality labs, and they're scrambling to join the artificial intelligence divisions. Because the AI teams are the ones receiving the blank checks for computing power and headcount. Exactly. Survival requires high visibility. Internal employee accounts actually reveal that workers are explicitly instructed to demand specific written feedback from managers about their promotion track. So this connects directly to what you experience when you open the app on your phone. If you are a quiet, heads down worker, like the silent rinder maintaining the basic infrastructure or working on trust and safety, you get fired. Yeah, the loud employee who aggressively aligns with the new AI mandates gets retained. So the bugs do not get fixed because the engineer who used to fix them is trying to rebrand themselves as an AI specialist. The internal focus is entirely on the new mandate. I mean, the internal review system even factors in AI driven impact. It judges employees on how effectively they use internal automated tools to increase their output. If you are not utilizing the new tools, you are penalized. Wow. And this internal shift has direct external consequences. Meta is ending its third party fact checking program in the United States and replacing it with a crowdsourced community note system kind of similar to what X uses. The causal connection here is pretty direct. Meta is slashing its human workforce to fund enormous artificial intelligence data centers. Right. So to balance the books, they're offloading the expensive human intensive work of moderation onto the users themselves. Wait, let me make sure I understand the mechanics of this. They're replacing professional journalists and paid researchers with anonymous users on the internet. Yes. Researchers classify the act of writing and rating these notes as unpaid data labor extracted from the user base. Yeah. You're relying on individuals to perform complex verification work for free. That is wild. Historically, content moderation is incredibly expensive because humans suffer psychological tolls when reviewing difficult content. Exactly. And the platform is bypassing that cost entirely. But that opens up serious vulnerabilities. Trusting random users to moderate a global platform feels incredibly risky. It is. The oversight board explicitly warned about this. Relying on consensus from users is incredibly dangerous in countries with a history of coordinated disinformation networks. Because the community note system operates on a specific mathematical assumption, right? Yeah. It assumes a sufficiently diverse and independent set of contributors will evaluate content in good faith. What happens when the bad actors have more resources than the good faith users? Authoritarian regimes possess the technical sophistication to coordinate massive numbers of accounts. Right. So the algorithm calculates a score based on whether contributors who usually disagree with each other find a note helpful. It looks for a bridge between divided groups. But if a state-sponsored troll farm floods the system with coordinated accounts, they can artificially simulate that historical disagreement. Oh, I see. They can build fake profiles that appear to sit on opposite sides of the political spectrum and then have those accounts agree on a deceptive note. Exactly. And when they do that, the assumption of good faith consensus collapses. They can exploit the feature to manipulate the information ecosystem. They are essentially using the platform's new safety feature as a weapon. If a coordinated network downvotes accurate context or upvotes deceptive context, the algorithm is mathematically tweaked into believing a consensus has been reached. And that risk becomes vastly more acute as artificial intelligence facilitates the scaled creation and operation of these fake networks. Yeah. You no longer need human operators sitting in a warehouse to run these fake profiles. Yeah. The Oversight Board recommended that meta-o-mit countries with a historical pattern of intentional, large-scale disinformation networks from this program entirely. Right. Eventual inclusion should require rigorous testing, like red teaming program vulnerabilities, just to prove the safeguards actually work. But we are already seeing the friction points during active conflicts. During the Israel-Iran War, a completely deceptive AI-generated video showing extensive damage to buildings in Haifa received over 700,000 views on Facebook before being caught. With professional fact-checkers removed and crowdsourced notes vulnerable to manipulation, platforms are exposed to the exact artificial intelligence tools they are currently spending billions to develop. It creates a massive blind spot where deceptive output garners huge numbers of views in a soft war. Exactly. The algorithms that recommend content prioritized engagement and artificially generated conflict videos are perfectly designed to maximize that engagement. As a consequence, meta has been forced to apply AI info labels to manipulated media rather than deleting it. This limits the platform's burden regarding free speech moderation. Right. But it places the responsibility entirely on you, the listener, to discern reality. The Oversight Board actually recommended labeling as a less restrictive alternative to deletion. The premise is that not all manipulated media is harmful. Think of political satire or harmless parodies. The idea is to provide context without unduly restricting expression. Right. But I have to push back here. Is labeling actually better than deletion when the automated systems making these decisions are incredibly flawed? That is a fair question. I mean, Meta's algorithms once flagged a photograph of onions as sexual content. Oh yeah, that is a perfect example of how these moderation algorithms fail. They rely on things called word embeddings and image recognition patterns. Okay, what exactly is a word embedding? How does the algorithm confuse a vegetable with explicit material? Think of word embeddings as a multi-dimensional mathematical map of concepts. The algorithm plots words based on how often they appear together in training data. So it does not actually understand what an onion is. No, not at all. It just calculates distance between data points. If the word spicy or certain curved shapes are mapped closely to restricted content and the photo of the onion triggers those specific mathematical coordinates, the system fails. The data set used to train these models often reproduces strange associations. So we are trusting the math that banned an onion to accurately label sophisticated political deep fakes in a war zone. Which is exactly why researchers are exploring alternative methods to fix these inherent biases like paraphrasing technology. I was reading about this in the sources and frankly it sounds terrifying. It is a smart filter that automatically rewrites a user's hateful post into something polite before the recipient sees it. Instead of deleting the speech or labeling it, the system alters the semantic value of the message entirely. That is a real-time digital ventriloquist. That crosses an ethical line regarding speaker autonomy. A user's speech would be secretly altered by an algorithm without their knowledge. Imagine if you post a sarcastic criticism of an authoritarian regime and the smart filter decides your tone is too hostile. It rewrites your post to be polite, effectively making it look like you are endorsing the regime. You are completely distorting the author's intent to maintain a sanitized environment. Under international human rights law, restrictions on freedom of expression must be clearly articulated so that speakers know what the rules are. If a smart filter is silently rewriting your text, the interests protected by freedom of speech, like individual self-definition and democratic engagement, are seriously compromised. The legal frameworks we rely on assume that the words published under your name are actually the words you wrote. And if meta's automated systems are struggling to tell a difference between onions and deepfakes, and they are replacing human moderators with unpaid users, who is holding them accountable? Because depending on where you live, the government's answer is entirely contradictory. Global governments are stepping in to mandate how the internet should be policed. The European Union's Digital Services Act legally forces platforms to maintain strict moderation. While a federal appeals court in the United States upheld a Texas law prohibiting platforms from censoring user viewpoints. Exactly. That is like driving a car where the passenger in Europe is yanking the steering wheel left, and the passenger in Texas is yanking it right. This severely limits meta's ability to operate a unified global plan. Keep going! You're doing it! That's the sound of Sam learning to swim in a Hilton resort pool. Oh, that's delicious. And that's the sound of Sam and his family enjoying dinner in the hotel restaurant. Good evening. Welcome back. With stays in your favorite destinations and everything taken care of, you can savor what's important. When you want your holiday to feel like a holiday, it matters where you stay. Book now at hilton.com. Hilton for the stay. At EDF, we don't just encourage you to use less electricity. We actually reward you for it. That's why when you use less during peak times on weekdays, we give you free electricity on Sundays. How you use it is up to you. EDF. Change is in our power. Households and ship weekday peak usage by 40% for earn up to 16 hours of free electricity for subject to fare usage tax. For all season series, visit EDFenergy.com forward slash high power. In Europe, the Digital Services Act implements systemic risk assessments. It creates trusted flagger programs and allows users to legally challenge platforms through independent out of court dispute settlements. Let me clarify the trusted flagger program. This gives specific organizations a specialized status where their reports of illegal content are processed faster than a normal user's report. Correct. And platforms face immense financial penalties for non-compliance with the European framework, including fines of up to 6% of their global turnover. We are talking about billions of dollars in potential penalties. Yeah. Meanwhile, the Texas law treats social media platforms like common carriers, similar to a telephone company, restricting their ability to remove content based on viewpoint. A phone company just connects a wire, but a social network curates a feed. Those are completely different functions. If you treat a social network like a public utility, they lose the ability to stop spam or organized harassment because they have to serve everyone equally. The legal reasoning behind the Texas law argues that platforms merely engage in viewpoint based censorship with respect to expression they have already disseminated. They argue these companies are the modern public square. But this directly conflicts with European standards, where public incitement to hatred or violence is strictly outlawed, requiring proactive moderation. Right. They are legally required to leave content up in one jurisdiction and legally required to take it down in another. And when platforms face immense pressure to moderate, they quickly retreat when legally challenged. Meta tried to implement a PG-13 rating for accounts to signal safety. The intention was to borrow the cultural familiarity of a movie rating to show parents they were taking teen safety seriously. Why did they cave and abandon the idea? Well, the Motion Picture Association complained about the trademark. Trademark litigation is immediate and highly costly, whereas platform safety is abstract. It proves that despite creating these elaborate systems for safety, corporate policy is incredibly fragile when confronted with external legal friction. They will abandon a safety feature the moment it threatens their legal standing with another powerful corporation. Exactly. Let me summarize the reality of what we are looking at. Meta is funding its artificial intelligence ambitions by stripping away its human workforce, professional fact-checkers, and independent oversight. The systems designed to protect the user base are being automated or outsourced to the users themselves. It leaves you wondering, when the next global crisis floods the internet with targeted misinformation, will the remaining algorithms and unpaid users be enough to hold the line? Or are we entirely on our own? If you're not subscribed yet, take a second and hit follow on whatever app you're using. It helps us keep making this. We appreciate you being here.
