# Just went on CNN to discuss Anthropic, the persistent accuracy issues with the use of AI for targeting, and why we need to maintain baseline existing standards that wouldn’t allow these uses - the… **By:** 2,755 followers **Published:** 2026-02-27T21:49:16.192Z **Source:** [LinkedIn](https://www.linkedin.com/posts/sarahmyerswest_just-went-on-cnn-to-discuss-anthropic-the-ugcPost-7433266888044056578-7H05) --- [Sarah Myers West](https://www.linkedin.com/in/sarahmyerswest?trk=public_post_feed-actor-name) Just went on [CNN](https://www.linkedin.com/company/cnn?trk=public_post-text) to discuss Anthropic, the persistent accuracy issues with the use of AI for targeting, and why we need to maintain baseline existing standards that wouldn’t allow these uses - the lowest of possible bars! More on this here: [https://lnkd.in/gQdQvY88](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Flnkd%2Ein%2FgQdQvY88&urlhash=j3KZ&trk=public_post-text) ##### **Transcript** More let's bring in Sarah Myers, WA Co executive director at the AI Now Institute, who joins us live now from New York. Sarah, thank you so much for being with us. I just want to get your take just out of the gate. Just, you know, here you have the Defense Department, Pete Hegseth, essentially using DPA to sort of force a private AI company towards military objectives. Your thoughts on that? And also what does that say about autonomy, you know, general or sort of autonomy of private AI companies going forward? Thanks for having me. I mean, let's take a step back. The the Pentagon itself has rules that wouldn't allow the use of AI for autonomous weapons. There's a requirement that they be engaged in rigorous testing, that there be meaningful human control over these systems. And there's good reason for those rules to be upheld because tests of similar systems that have been made public show that the accuracy of using LM's in sort of battlefield environments is is extremely low. It's an unacceptable threshold for targeting is an unacceptable threshold for for the use of these systems for defense. And so, you know, this though this is being framed as an issue around a particular companies red lines. I think the key here is that there are existing standards around the use of tools in these contexts that need to be upheld in that the Pentagon should be abiding by its own playbook here. Yeah, that's an interesting point that it's not just about anthropic, it's really about the the Pentagon sort of upholding their own their own rules on this. But from anthropic perspective, you know, talk to us about the ethical considerations. I mean, where do they draw the line between national security interests versus, you know, the use of that technology for mass surveillance, for autonomous weapons? I mean, ethically, you know, they've made their position very clear, but explain that thought process to us that they're in. Yes, I mean, I think it's, it's wise that Anthropic has put safety forward in its, you know, orientation towards these tools. You know, if they are to be deployed in really highly, you know, socially significant contacts like surveillance or or weaponry that would have profound implications for the public at large. But we've seen Anthropic itself feel immense pressure to roll these guidelines back. For example, they release they recently rolled back. They're responsible scaling policy, which said that, you know, that if they found that the systems that they were developing outstripped their ability to control them, they would stop building them. And they've, they've recently removed this policy because none of their competitors are keeping up with it. They say themselves they're in a race to the bottom and because they're seeing so much political pressure more generally around safety. So I think the entire context around AI and the the safety of these tools right now is in is in a quite. Higher student, it's really important for policymakers to be, you know, pushing the the line for baseline protections for for the public from these, you know, quite egregious potential uses. I want you to expand on something you you sort of touched on a minute ago. What would have been the consequences would have been the broader implications had Anthropic acquiesced to to the Department of Defense demands here, Pete Hegseth demands. What we've seen in recent memos coming from the Pentagon that there are particular kinds of uses of AI that they really want to be doubling down on. It's things like enabling AI to be used autonomously in drone warfare, The use of AI agents that can act, you know, without human oversight on the development of, of drone swarms, mass math that sort of potentially uses not by the Pentagon, but by DHS for mass surveillance. These are things that, you know, historically that the United States has not exceeded to and it's really like a massive ramp up in the militarization of these tools. And final question, could this standoff prompt Congress to sort of rewrite laws or just regulators to really think rethink rather AI policy going forward? Let's hope so. I mean, I think that there's, there's already ground to just focus on upholding the rules that we already have. You know, we don't need to rewrite the playbook for this, this version of AI, their existing safety standards that we've used across many different generations of technology that have really served us well, that protected our security both at the national level and the security of of the, you know, the domestic public. So starting there just by making sure that we're upholding and enforcing. The existing rules is is a really good start and unfortunately it seems like that's, you know, that's the the lowest bar that we can we can be moving from, right. Sarah Myers W thank you so much. They moved on and secured a deal with OpenAI, allegedly including the same ethical safeguards that Anthropic had requested to include, read statement from Dario Amodei: [https://www.politico.com/news/2026/02/28/openai-announces-new-deal-with-pentagon-including-ethical-safeguards-00805546](https://www.politico.com/news/2026/02/28/openai-announces-new-deal-with-pentagon-including-ethical-safeguards-00805546?trk=public_post_comment-text) This is why Ethics is a discipline. How long before they manufacture a war to test? This is my current focus. What are your thoughts on preparing for this “In 2025, Google’s Willow chip achieved verifiable quantum advantage by running an algorithm 13,000 times faster than the most powerful supercomputers, a task that would have taken a classical supercomputer 10 septillion years.” Both Google and IBM Quantum Computers solved similar problems in 5mins. IBM Quantum Computing forecast is 2029 [https://www.ibm.com/quantum/blog/large-scale-ftqc](https://www.ibm.com/quantum/blog/large-scale-ftqc?trk=public_post_comment-text) Yah Sarah! FYI my friend [Dunstan Allison-Hope](https://www.linkedin.com/in/dunstanallisonhope?trk=public_post_comment-text) just wrote a thing that is relevant here. [https://www.linkedin.com/pulse/ai-brings-intelligence-wisdom-why-anthropic-needs-allison-hope-akp2c](https://www.linkedin.com/pulse/ai-brings-intelligence-wisdom-why-anthropic-needs-allison-hope-akp2c?trk=public_post_comment-text)/ What do you think to this [Sarah Myers West](https://www.linkedin.com/in/sarahmyerswest?trk=public_post_comment-text)?[republic.theailyceum.com](http://republic.theailyceum.com?trk=public_post_comment-text) Thanks [Sarah Myers West](https://www.linkedin.com/in/sarahmyerswest?trk=public_post_comment-text) for settling the record straight: if the Pentagon’sown rules prohibit LLM use in autonomous weapons, this whole standoff was for nothing. & now watch Altman trying to get some virtue washing points… Important conversation to bring to a mainstream stage. When [CNN](https://www.linkedin.com/company/cnn?trk=public_post_comment-text) is spotlighting [Anthropic](https://www.linkedin.com/company/anthropicresearch?trk=public_post_comment-text) and targeting accuracy, it signals this is no longer a niche ethics debate—it’s a real-world risk question. “Lowest bar” still means real standards. Speed without guardrails isn’t innovation; it’s exposure. [Sarah Myers West](https://www.linkedin.com/in/sarahmyerswest?trk=public_post_comment-text), [Frances West](https://www.linkedin.com/in/frances-west-0b81924?trk=public_post_comment-text), [Private Directors Association®- New England](https://www.linkedin.com/showcase/pdanewengland/?trk=public_post_comment-text) [Sarah Myers West](https://www.linkedin.com/in/sarahmyerswest?trk=public_post_comment-text) rightly reminds us that the DoD has its OWN RULES on safety thresholds - it needs to stick to those.