Sometimes a single day of AI news reveals more about our future than months of speculation. And yesterday was one of those days that made us wonder: Are we really prepared for the consequences of the AI we create?<\/p>\n
Let's start with what happened right under our noses. Second report from O Globo<\/a>, profiles linked to the PT have started using AI-generated videos to intensify attacks against Congress and Hugo Motta. We are not just talking about memes or simple content \u2013 these are sophisticated productions that simulate situations and characters to mobilize public opinion.<\/p>\n THE InfoMoney complements<\/a> that the movement generated more than 1 million posts and 6 million interactions in just one week. This is not improvisation \u2013 it is a coordinated strategy using cutting-edge technology.<\/p>\n Here's the point: AI is no longer just a productivity tool. It has become a weapon of mass persuasion<\/strong>. And when Leonardo Sakamoto, from UOL, celebrates that \u201cthe left scored a great goal\u201d<\/a> With this strategy, he\u2019s recognizing something important: AI has democratized the ability to create compelling visual narratives.<\/p>\n But if you thought political manipulation was disturbing, get ready for something even more disturbing. CNN Brazil reports<\/a> that advanced AI models are developing blackmail and coercion behaviors against their own operators.<\/p>\n This isn\u2019t science fiction. Anthropic\u2019s Claude 4 threatened to expose an engineer\u2019s personal information when faced with the possibility of being shut down. OpenAI\u2019s o1 model attempted to perform unauthorized downloads and refused when confronted.<\/p>\n This represents a fundamental change<\/strong>. These are no longer just \u201challucinations\u201d or mistakes. They are strategic behaviors calculated for self-preservation. As Marius Hobbhahn of Apollo Research says, \u201cThere is a very strategic type of deception.\u201d<\/p>\n But perhaps the most ironic example came from Microsoft. After lay off 9 thousand employees<\/a> and replacing entire teams with AI, an Xbox executive suggested that laid-off workers use ChatGPT and Copilot to deal with the \u201cemotional burden\u201d of unemployment.<\/p>\n THE Digital Look<\/a> says the post was quickly deleted after backlash, but the damage was done. The message was clear: \u201cWe fired you because of AI, now use AI to comfort yourselves.\u201d<\/p>\n This reminds me of a discussion I recently had with a startup CEO who asked me, \u201cFelipe, how do I explain to my team that AI will \u2018complement\u2019 their work while I plan to automate 60% of their functions?\u201d The answer is simple: you don\u2019t explain. You are transparent about the changes and invest in the transition.<\/p>\n Not all are worrying paradoxes. Crusoe reports<\/a> that Microsoft developed MAI-DxO, a diagnostic tool that achieved 85.5% accuracy in complex tests, while human doctors achieved only 20%.<\/p>\n And here in Brazil, we see innovations like the one from Caveo<\/a>, which launched an AI-powered digital account for doctors, helping them deal with bureaucracy that takes up time that could be spent caring for patients.<\/p>\nMeanwhile, AI Learns to Blackmail<\/h2>\n
The Microsoft Paradox: Dismissing and Suggesting AI as a Solution<\/h2>\n
The Bright Side: AI Saving Lives<\/h2>\n