This data is reliable.
26.09.2025 12:59 β π 5 π 1 π¬ 3 π 1@brianfishman1.bsky.social
National security, technology, California. Cinder co-founder, led DOI at FB, created Palantir disaster relief, Research Director at CTC @ West Point.
This data is reliable.
26.09.2025 12:59 β π 5 π 1 π¬ 3 π 1I know Bluesky is still not foreign policy / national security central, but for those of us involved in CT land, this is some mind-blowing stuff right here.
Lots of ISIS dudes rolling over in their graves right now.
Corruption.
www.msnbc.com/msnbc/news/t...
Awful and unacceptable.
www.denverpost.com/2025/09/11/e...
Political violence is unacceptable. Full stop.
10.09.2025 21:12 β π 4 π 0 π¬ 0 π 0#infosec
10.09.2025 16:39 β π 1 π 0 π¬ 0 π 0What a shitshow.
www.bbc.com/news/article...
I have deep respect for West Point, having taught there for four years. These days, they are not living up to their principles.
06.09.2025 19:45 β π 7 π 1 π¬ 0 π 0If you don't test your AI at the bar, does it even count?
20.08.2025 18:57 β π 1 π 0 π¬ 0 π 0Since security guarantees are back in discussion, for those interested in the requirements, tasks, and what a European deterrence force might look like - check out this piece from back in March with Jack Watling. warontherocks.com/2025/03/will...
17.08.2025 15:27 β π 174 π 36 π¬ 3 π 1Awful.
12.08.2025 14:24 β π 10 π 3 π¬ 0 π 0Yes and no. Sheβs on to something, but βStop the Stealβ was more than an algorithm game (which is oversimplified in that article). You canβt overstate the impact of the President legitimizing an emotionally-resonant - we didnβt really lose!!! - conspiracy theory from the Oval.
10.08.2025 15:35 β π 3 π 0 π¬ 0 π 0I had the luck to edit this back in the day. Lots of really valuable information in here - definitely worth the read.
07.08.2025 18:56 β π 4 π 2 π¬ 1 π 0I appreciate journalists trying to cover the general trend here. But I suspect a misinformed public here is a function of the proliferation of media sources & coverage coupled w/ very old βif it bleeds it leadsβ media dynamics. Hard to beat plethora of legit anecdotes w a few data-driven stories.
31.07.2025 02:41 β π 9 π 0 π¬ 0 π 0This undermines U.S. national security. I taught in SOSH & was always proud of USMA's willingness to stand up to political pressure. The institution protected our independence and thereby the quality of the education we provided cadets. Politicizing education and our military is garbage.
30.07.2025 18:58 β π 25 π 13 π¬ 0 π 0I donβt know if @jmberger.com will be right or wrong, but heβs thought about this as much as anyone.
26.07.2025 23:33 β π 2 π 0 π¬ 1 π 0This is a critical story, with implications far beyond Ukraine.
www.reuters.com/investigatio...
In the wake of the Texas floods, some self-described veterans have seemingly lost their minds and are attacking weather radars. But a reminder that there are far more veterans continuing their service in this moment.
teamrubiconusa.org/news-and-sto...
Monarchies are authoritarian states with better branding.
10.07.2025 00:08 β π 10 π 1 π¬ 0 π 0Turns out AI Safety matters. Cinder's customers - you've heard of them - use our product to do stuff like: remove CSAM, identify nation-states, prevent customer service chatbots from going nazi on diverse customer bases, comply with sanctions law. Almost like its just good, basic business practice.
08.07.2025 23:49 β π 6 π 1 π¬ 0 π 0I have questions.
08.07.2025 00:24 β π 2 π 1 π¬ 2 π 0Also, are you sure you're going to have enough to talk about?
23.06.2025 21:21 β π 0 π 0 π¬ 1 π 0This better be good, Afshon.
23.06.2025 21:20 β π 2 π 0 π¬ 2 π 0Anyway, I think thatβs the context here. And, fwiw, I give the model developers credit for highlighting risks. Sure, some of it is disaster porn that generates clicks. But itβs also a much better effort than social media in its early days. Perfect? No. But positive trend. 8/8
20.06.2025 19:30 β π 0 π 0 π¬ 0 π 0Think this model is too favorable to foundation model developers? Of course the model developers are going to come up w/ a vision of responsibility that aligns to their business interests. But I do think they are laying out that vision and I donβt hear a lot of thoughtful examination. 7/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0Fwiw, Iβm not necessarily endorsing this breakdown of responsibility, though I agree that the foundation models have to address the catastrophic risks. But I do think this allocation of responsibility is the first key issue. 6/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0The app developers also donβt really have the focus or means (in many cases) to manage the security issues. But because they are close to the usecase and the user end users they do have relevant info that the model developers donβt necessarily have. 5/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0That doesnβt mean they arenβt working on other AI safety issues. They are. But it does mean that they are most aggressively accepting responsibility for the catastrophic risks. The issue here is that app developers do not really understand this breakdown of responsibility. 4/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0Itβs not the cost of content moderation that they worry about most; itβs about honing a model thatβs unlikely to produce harm but that is still applicable to a wide range of usecases. Might sound easy, but it isnβt. So you see companies focusing on catastrophic risks, like bioweapons. 3/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0This will push more responsibility on app developers to manage lesser (but more common and still highly problematic) risks. Itβs fair to say that foundation model companies see this as financially useful, but that doesnβt mean itβs necessarily wrong. 2/n
20.06.2025 19:30 β π 0 π 0 π¬ 1 π 0