<div dir="ltr"><div class="gmail_quote gmail_quote_container"><div dir="ltr" class="gmail_attr">On Sat, 28 Feb 2026 at 22:42, spike jones via extropy-chat <<a href="mailto:extropy-chat@lists.extropy.org">extropy-chat@lists.extropy.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
There is no going back. If the bad guy is developing AI weapons, then AI weapons are required for defense against them. It is race to the technology or die. The Chinese get it.<br>
This isn't that new really. The navy has had loitering target-identification torpedoes for some time.<br>
<br>
Regarding what the pentagon is doing with AI, it is perfectly reasonable to strictly control what goes into the AI's training. The generals must know the AI-enabled weapon they carry into battle will not turn around and kill the soldiers who fired it.<br>
<br>
spike<br>
_______________________________________________<br></blockquote><div><br></div><div><br></div><div><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">I wonder whether Hegseth has considered asking an AI about this policy? </div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">I just did, and Perplexity AI described it as a race to the bottom.</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">BillK</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">Perplexity AI -</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><br></div></div><div class="gmail_quote gmail_quote_container"><p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">The
policy you describe would push military AI toward maximized capability
and speed at the expense of safety, civil liberties, and long‑term
strategic stability. If other major powers follow the same path, it
risks a global “race to the bottom” on AI guardrails, especially in
surveillance and weapons systems.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:73px"></span></span></p>
<h2 class="gmail-mb-2 gmail-mt-4 gmail-[.has-inline-images_&]:clear-end gmail-font-editorial gmail-font-bold gmail-text-lg gmail-first:mt-0 gmail-md:text-lg gmail-[hr+&]:mt-4" id="gmail-what-the-pentagon-is-pushing-for">What the Pentagon is pushing for</h2>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Recent
reporting describes U.S. defense officials pressuring Anthropic to
allow their models to be used for any “lawful” purpose, explicitly
including AI‑directed weapons and domestic surveillance, and threatening
to end a large contract if the company refuses. Anthropic, by contrast,
has drawn firm lines against mass domestic surveillance and fully
autonomous weapons, even while supporting many other national‑security
uses like intelligence analysis and cyber defense.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:37px"></span></span><span class="gmail-citation-nbsp"></span><span class="gmail-inline-flex" aria-label="Pentagon vs Anthropic: The Battle Over AI Guardrails | Vantage with Palki Sharma | N18G"><span class="gmail-citation gmail-inline" rel="nofollow noopener"></span></span></p>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">The
Defense Department’s stance is that internal corporate safety limits
are “obstacles” if the underlying use is legal, and that the military
must retain full control over how AI is deployed in war and intelligence
operations. This pits operational flexibility and secrecy against
vendor‑imposed ethical and legal constraints.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
<h2 class="gmail-mb-2 gmail-mt-4 gmail-[.has-inline-images_&]:clear-end gmail-font-editorial gmail-font-bold gmail-text-lg gmail-first:mt-0 gmail-md:text-lg gmail-[hr+&]:mt-4" id="gmail-domestic-implications-surveillance-rights-and-powe">Domestic implications: surveillance, rights, and power</h2>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">If
“no guardrails” effectively means allowing broad AI‑enabled
surveillance so long as some legal justification can be found, several
consequences follow:</p>
<ul class="gmail-marker:text-quiet gmail-list-disc"><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2"><strong>Mass domestic surveillance at scale</strong>:
AI can aggregate data from email, browsing history, movements, and
other purchased or intercepted data to build detailed profiles of
citizens, turning fragmented, controversial practices into a coherent,
far more invasive capability. This is materially different from
human‑scale monitoring because of its reach, persistence, and ability to
infer sensitive attributes.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Weakening
of civil liberties: Current law already lags behind technology, and
many AI‑enabled surveillance tactics are not clearly illegal even if
they undermine privacy and freedom of association. Once such tools exist
and are normalized, they are hard to roll back and can chill protests,
journalism, and political opposition.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:91px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Reduced
accountability: If models are used for watchlists, risk scores, or
“pattern of life” analysis without strict constraints and audit trails,
their outputs can become de facto targeting or profiling decisions that
are opaque and difficult to challenge. Errors or bias then have
real‑world consequences but no clear line of responsibility.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
</li></ul>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">In
short, removing guardrails in this domain risks moving liberal
democracies toward the kind of pervasive, data‑driven social control
they often criticize in authoritarian systems.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
<h2 class="gmail-mb-2 gmail-mt-4 gmail-[.has-inline-images_&]:clear-end gmail-font-editorial gmail-font-bold gmail-text-lg gmail-first:mt-0 gmail-md:text-lg gmail-[hr+&]:mt-4" id="gmail-battlefield-and-escalation-risks">Battlefield and escalation risks</h2>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">On the military side, the key danger is sliding from AI “decision support” into de facto or full autonomy in lethal force:</p>
<ul class="gmail-marker:text-quiet gmail-list-disc"><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">From
support to de facto control: AI systems already aggregate sensor data,
predict enemy actions, and support targeting decisions; without
guardrails and human‑in‑the‑loop requirements, commanders may come to
rely on recommendations that function as effective orders. Under time
pressure, the human “check” can become ceremonial.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Fully
autonomous weapons: Anthropic and many legal scholars explicitly warn
that weapons that identify, select, and engage targets without
meaningful human control pose serious ethical and legal problems, even
if not yet clearly banned. They complicate compliance with distinction
and proportionality in the laws of war, and make attribution and
accountability for war crimes harder.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Escalation
and accident risk: High‑speed, opaque systems interacting in contested
environments (e.g., drone swarms, automated cyber responses) increase
the chance of misinterpretation, unintended engagements, and rapid
escalation before humans can intervene. Once multiple states adopt such
systems, crisis stability becomes more fragile.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
</li></ul>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">A
“no guardrails” posture thus prioritizes short‑term tactical advantage
over long‑term control of escalation and adherence to international
humanitarian law.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
<h2 class="gmail-mb-2 gmail-mt-4 gmail-[.has-inline-images_&]:clear-end gmail-font-editorial gmail-font-bold gmail-text-lg gmail-first:mt-0 gmail-md:text-lg gmail-[hr+&]:mt-4" id="gmail-international-race-dynamics-china-russia-and-other">International race dynamics: China, Russia, and others</h2>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">China
and Russia already view AI as central to military modernization,
information dominance, and intelligence, surveillance, and
reconnaissance (ISR). China leverages its vast data resources and
civil‑military fusion to build powerful surveillance and command
systems; Russia emphasizes autonomous systems, robotics, and AI‑assisted
reconnaissance and targeting.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">If the U.S. effectively declares that operational needs trump safety guardrails, several dynamics are likely:</p>
<ul class="gmail-marker:text-quiet gmail-list-disc"><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Race
to the bottom: Each side can point to others’ aggressive use of AI to
justify loosening its own constraints, arguing that restraint would be a
unilateral disadvantage. This discourages meaningful human‑control
requirements or surveillance limits.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Norms
and governance erosion: Efforts at international regulation of lethal
autonomous weapons and surveillance norms rely on leading powers showing
some willingness to accept constraints. If the U.S. insists on
unconstrained lawful use, it weakens its ability to argue against
similarly unconstrained systems deployed by China, Russia, or others.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:91px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Export
and proliferation: Once unrestricted military AI systems exist, they
will diffuse via arms sales, proxies, and commercial adaptation, making
powerful surveillance and targeting tools available to less accountable
regimes. That amplifies repression and instability globally.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p>
</li></ul>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">So
the policy is not just about U.S. capability; it helps set the global
standard for what is seen as acceptable military and security use of AI.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:91px"></span></span></p>
<h2 class="gmail-mb-2 gmail-mt-4 gmail-[.has-inline-images_&]:clear-end gmail-font-editorial gmail-font-bold gmail-text-lg gmail-first:mt-0 gmail-md:text-lg gmail-[hr+&]:mt-4" id="gmail-governance-and-policy-choices">Governance and policy choices</h2>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Analysts
describe the current clash over guardrails as a governance “inflection
point” for how frontier AI will be integrated into defense and
intelligence institutions. Several policy choices are in play:<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
<ul class="gmail-marker:text-quiet gmail-list-disc"><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Who
sets the limits: One model gives governments near‑total discretion,
with companies treated as arms‑length vendors; another allows developers
to embed legal and ethical restrictions into their products, even for
state clients. The outcome will shape procurement standards and the
balance of power between states and major AI firms.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:73px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Embedding
accountability: Strong guardrails would include clear scope
limitations, logging and auditability, mandatory human authorization for
high‑risk uses, and mechanisms to trace specific decisions back to
responsible commands. A “no guardrails” approach tends to undercut these
structures in favor of operational freedom.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:97px"></span></span></p>
</li><li class="gmail-py-0 gmail-my-0 gmail-prose-p:pt-0 gmail-prose-p:mb-2 gmail-prose-p:my-0 gmail-[&>p]:pt-0 gmail-[&>p]:mb-2 gmail-[&>p]:my-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Aligning
law with capability: Much of what is technically possible is “lawful”
simply because legal frameworks have not yet caught up. Without
proactive regulation—on surveillance thresholds, data use, and
autonomous weapons—“lawful use” becomes a very weak safeguard.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:73px"></span></span></p>
</li></ul>
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">In
effect, the Pentagon’s position asserts that legality plus internal
military processes are enough, while critics argue that for systems this
powerful and opaque, <strong>ex ante</strong> technical and contractual constraints are essential complements to law and doctrine.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:73px"></span></span></p>
<hr class="gmail-bg-quiet gmail-h-px gmail-border-0">
<p class="gmail-my-2 gmail-[&+p]:mt-4 gmail-[&_strong:has(+br)]:inline-block gmail-[&_strong:has(+br)]:pb-2">Taken
together, a push for AI “without guardrails” in surveillance and
military contexts risks entrenching mass monitoring, eroding civil
liberties, increasing the chance of accidental or unlawful use of force,
and driving a global competition toward ever less constrained systems. A
more cautious approach would insist on meaningful human control, narrow
and auditable uses, and international norms that apply equally to the
U.S., China, Russia, and others.<span class="gmail-citation-nbsp"></span><span class="gmail-group/trigger gmail-inline-flex gmail-min-w-0"><span class="gmail-citation gmail-inline-flex gmail-min-w-0" style="width:85px"></span></span></p><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">-----------------------------------------------------</div><br></div></div>