⸻ Podcast: Redefining Society and Technologyhttps://redefiningsocietyandtechnologypodcast.com _____________________________This Episode’s SponsorsBlackCloak provides concierge cybersecurity protection to corporate executives and high-net-worth individuals to protect against hacking, reputational loss, financial loss, and the impacts of a corporate data breach.BlackCloak: https://itspm.ag/itspbcweb_____________________________A Musing On Society & Technology Newsletter Written By Marco Ciappelli | Read by TAPE3August 9, 2025The Agentic AI Myth in Cybersecurity and the Humanity We Risk When We Stop Deciding for OurselvesReflections from Black Hat USA 2025 on the Latest Tech Salvation NarrativeWalking the floors of Black Hat USA 2025 for what must be the 10th or 11th time as accredited media—honestly, I've stopped counting—I found myself witnessing a familiar theater. The same performance we've seen play out repeatedly in cybersecurity: the emergence of a new technological messiah promising to solve all our problems. This year's savior? Agentic AI.The buzzword echoes through every booth, every presentation, every vendor pitch. Promises of automating 90% of security operations, platforms for autonomous threat detection, agents that can investigate novel alerts without human intervention. The marketing materials speak of artificial intelligence that will finally free us from the burden of thinking, deciding, and taking responsibility.It's Talos all over again.In Greek mythology, Hephaestus forged Talos, a bronze giant tasked with patrolling Crete's shores, hurling boulders at invaders without human intervention. Like contemporary AI, Talos was built to serve specific human ends—security, order, and control—and his value was determined by his ability to execute these ends flawlessly. The parallels to today's agentic AI promises are striking: autonomous patrol, threat detection, automated response. Same story, different millennium.But here's what the ancient Greeks understood that we seem to have forgotten: every artificial creation, no matter how sophisticated, carries within it the seeds of its own limitations and potential dangers.Industry observers noted over a hundred announcements promoting new agentic AI applications, platforms or services at the conference. That's more than one AI agent announcement per hour. The marketing departments have clearly been busy.But here's what baffles me: why do we need to lie to sell cybersecurity? You can give away t-shirts, dress up as comic book superheroes with your logo slapped on their chests, distribute branded board games, and pretend to be a sports team all day long—that's just trade show theater, and everyone knows it. But when marketing pushes past the limits of what's even believable, when they make claims so grandiose that their own engineers can't explain them, something deeper is broken.If marketing departments think CISOs are buying these lies, they have another thing coming. These are people who live with the consequences of failed security implementations, who get fired when breaches happen, who understand the difference between marketing magic and operational reality. They've seen enough "revolutionary" solutions fail to know that if something sounds too good to be true, it probably is.Yet the charade continues, year after year, vendor after vendor. The real question isn't whether the technology works—it's why an industry built on managing risk has become so comfortable with the risk of overselling its own capabilities. Something troubling emerges when you move beyond the glossy booth presentations and actually talk to the people implementing these systems. Engineers struggle to explain exactly how their AI makes decisions. Security leaders warn that artificial intelligence might become the next insider threat, as organizations grow comfortable trusting systems they don't fully understand, checking their output less and less over time.When the people building these systems warn us about trusting them too much, shouldn't we listen?This isn't the first time humanity has grappled with the allure and danger of artificial beings making decisions for us. Mary Shelley's Frankenstein, published in 1818, explored the hubris of creating life—and intelligence—without fully understanding the consequences. The novel raises the same question we face today: what are humans allowed to do with this forbidden power of creation? The question becomes more pressing when we consider what we're actually delegating to these artificial agents. It's no longer just pattern recognition or data processing—we're talking about autonomous decision-making in critical security scenarios. Conference presentations showcased significant improvements in proactive defense measures, but at what cost to human agency and understanding?Here's where the conversation jumps from cybersecurity to something far more fundamental: what are we here for if not to think, evaluate...
続きを読む
一部表示
6 分
27 分
17 分
12 分