Page 5

^What are you suggesting? I know what the letters stand for, if that's what you're referring to

2/26/2026 4:51:53 PM

But are you black?

2/26/2026 5:57:02 PM

I have definitely said FAQ phonetically back in the day

[Edited on February 26, 2026 at 7:16 PM. Reason : lemme go find an animated Gee Eye Eff file to post]

2/26/2026 7:16:13 PM

2/26/2026 8:35:59 PM

I'll give plus points to anthropic for an act of resistance

2/27/2026 5:17:10 PM

Quote :

"Anthropic’s contract with the Pentagon is worth as much as $200 million. The company had sought to have protections written into the deal against the use of its models for autonomous weapons and mass domestic surveillance."

I don't get the big deal about the bolded part. Like as much as the NSA spies on us, collecting phone records, metadata, etc...what could AI possibly do thats not already done now?

2/27/2026 9:08:27 PM

ask AI

2/27/2026 10:24:36 PM

^^The concern about AI and mass surveillance isn’t just that governments collect large amounts of data — agencies like the National Security Agency have done that for years. The key difference is what AI allows them to do with that data.

In the past, massive data collection still required large teams of human analysts to interpret it. That created natural limits. AI removes those bottlenecks. It can automatically analyze millions of conversations, detect patterns, infer political leanings or emotional tone, map social networks, and flag “suspicious” behavior in real time. Instead of just recording what people did, AI can model and predict what they might do.

That shift—from passive storage to automated interpretation and prediction at population scale—is what raises civil liberties concerns. It lowers the cost and friction of monitoring everyone, making broad, always-on surveillance technically feasible.

Companies like Anthropic worry that their models could enable automated repression or large-scale behavioral scoring systems, similar to concerns often raised about systems in China. The debate isn’t about whether surveillance exists today—it’s about how dramatically AI could expand its power and reach tomorrow.





I asked(in around 100 words) "whats the problem with always-on surveillance because if anything it seems like a good thing...it deters crime, terrorism, etc

Quote :

"Supporters of broad, always-on surveillance argue it increases safety, deters crime, and helps prevent threats before they happen. If someone isn’t doing anything wrong, they believe there’s little to fear. Critics argue the problem isn’t just spying — it’s power imbalance and long-term risk. When authorities can monitor everyone, people may self-censor, reducing genuine freedom. AI systems can also make errors, falsely flagging individuals at scale. Surveillance programs often expand beyond their original purpose, and stored data can later be reinterpreted under new laws or leadership. The core debate is whether increased security is worth sacrificing private, unmonitored space."

I'm a strong proponent of "If someone isn’t doing anything wrong, they believe there’s little to fear."

2/27/2026 10:40:24 PM

^ Do you want AI mistaking your account for terrorist activity and stealing your life savings?

Think SWATing instances where the police barge into the wrong home and blow some dude away because they got the address wrong. Do you want a model that hallucinates responsible for determining if you get detained for 12 hours? Even if you have nothing to hide they can freeze your accounts by mistake or deny all kinds of services that you may use and need. Including banking and credit. There have been countless cases of people going to buy cars with cash and getting it swiped by the government or police at a stop. Any interaction with gov has the risk of a mistake happening.

Usually when the gov fucks up massively it becomes your responsibility to prove it. If you aren't a pretty girl or some aggrieved class the chances of you winning or getting anyone to care are low.

Say the mistake rate for targeting the wrong person is 0.5% and with current methods they wrongly accuse and harass 5000 people. With new AI tools they cover say 20x as many people. Assuming the mistake rate is the same that's an extra 95,000 people that are going to be wrongly harassed.





[Edited on February 28, 2026 at 7:32 AM. Reason : It's estimated between 1 and 5% of convictions are wrongful]

2/28/2026 7:20:56 AM