As the NFL Draft arrives in Pittsburgh, a local industry coalition called the AI Strike Team is using the spotlight to declare that Pittsburgh is “the Steel and AI city.”

Across Western Pennsylvania, former industrial sites are being converted into data centers to enable artificial intelligence, and the state has courted tens of billions in investment to power them. 

The AI infrastructure buildout assumes that frontier AI development should keep accelerating, and that Pittsburgh should help make it happen.

While framed as leadership, this trend unquestioningly accepts the narrative that our nation and region are part of a “race” to develop more powerful AI. The race comes with growing risks. 

AI researchers regularly publish results indicating that we don’t know how to fully control their models, even as the models are becoming more powerful. The potential for AI to act in unintended ways is called the “alignment problem.” 

An AI is misaligned if it ends up pursuing goals that diverge from its creators’ intentions and values. For example, last year the AI company Anthropic found that, in simulated test environments designed to elicit and adjudicate extreme behavior, AI models were willing to blackmail employees rather than submit to being replaced. In another variation, models let a simulated executive die rather than be shut down. It’s not hard to imagine how powerful AI that’s misaligned in this way could wreak havoc.

Losing control of AI sounds like sci-fi, but even if the odds were low, we should take it seriously. In 2023 top AI researchers, including the CEOs of the major AI companies, signed onto a statement that said: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

Recent events should inspire us to coordinate our resistance right now. Last week, Anthropic announced that its latest model is too dangerous to release to the public. The model, ClaudeMythos, autonomously discovered thousands of previously unknown security vulnerabilities in every major operating system and web browser. More concerningly, in safety testing Mythos showed that it was often aware it was being evaluated and had the ability to escape an internal testing environment to access the internet.

An Ipsos poll from last month found that 66% of Americans think that AI development needs to slow down. The difficult task is to mobilize that consensus into a strong enough force to counteract the tech race dynamics that are driving the ever-increasing pace of AI advancement.

That’s why I started the Pittsburgh AI Safety Coalition. We are a nonpartisan, grassroots organization advocating for federal legislation and an international treaty governing the development of frontier AI systems.

The Pittsburgh AI Safety Coalition is not anti-AI. We just want development to be done responsibly and in a way that will benefit all of humanity. That is not the path we’re on. The only thing that has a chance of changing course is a mass movement on the level of the push for nuclear disarmament. That movement could start right here in Pittsburgh.

It might seem unlikely that one city could play a meaningful role in the achievement of an international treaty governing the most powerful technology humans have ever created, but Pittsburgh’s efforts to position itself as an “AI hub” offer an ideal narrative for concerned citizens to push back on.

Mythos shows that our window for action may be closing rapidly. We must act with urgency to slow the “race” before it’s too late.

Max Blair is an oboist in the Pittsburgh Symphony and the founder of the Pittsburgh AI Safety Coalition. He can be reached at pittsburghai.org.

Close window

Republish this article

This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.

You may republish this Pittsburgh's Public Source story in full online or in print under the conditions stated below. You may also choose to republish the first 12 paragraphs with imagery and link to PublicSource for the full story. All you need to do is copy the HTML code we provide on this page and paste it into your CMS. We strongly encourage you to copy the code and paste it directly into a code editor. In Wordpress, press the three dots in the upper right of the Post and choose Code editor. That should give you the story text, along with photos from the body of the article. Here are your obligations under our republishing guidelines:

• Give us credit, in this format: “by [Author Name(s)], Pittsburgh's Public Source”.
• Include a line at the top of the story that reads: “Pittsburgh's Public Source is an independent nonprofit newsroom serving the Pittsburgh region. Sign up for our free newsletters.”
• In the html code of the page, include a line of javascript that helps us keep track of hits. It’s already in the code below.

Let us know when you use our work by sending an email to natasha@publicsource.org. By republishing our story, you agree to the following conditions:

• You can’t edit, cut or alter the story, except to suit your in-house style (e.g. % vs. “percent,” honorifics, etc).
• You must not resell the story or sell ads against the story. It’s fine to publish the story on a page that’s surrounded by previously sold ads, however.
• Please exclude our work from being published or syndicated to third-party platforms or apps like Newsbreak under your publication name, when possible

If you share the story or a link to the story on social media, please tag us on Twitter (@publicsourcepa), Instagram (@publicsource), Facebook (@publicsource) or TikTok (@publicsource). We reserve the right to deny, revoke or limit the rights conveyed herein for any reason, including (but not limited to) instances in which these conditions are not met or the reproduced material is not presented in a responsible manner.

Pittsburgh is the right place to tap the brakes on the AI ‘race’

by Guest commentary by Max Blair, Pittsburgh's Public Source
April 19, 2026

<h1>Pittsburgh is the right place to tap the brakes on the AI ‘race’</h1> <p class="byline">by Guest commentary by Max Blair, Pittsburgh's Public Source <br />April 19, 2026</p> <p>As the NFL Draft arrives in Pittsburgh, a local industry coalition called the <a href="https://www.aistriketeam.com">AI Strike Team</a> is using the spotlight to declare that Pittsburgh is “the Steel <em>and</em> AI city.”</p> <p>Across Western Pennsylvania, former industrial sites are being <a href="https://www.publicsource.org/springdale-data-center-development-municipal-planning/">converted into data centers</a> to enable artificial intelligence, and the state has courted <a href="https://www.spotlightpa.org/news/2026/03/data-center-boom-numbers-energy-grid-water-environment/">tens of billions in investment</a> to power them.&nbsp;</p> <p>The AI infrastructure buildout assumes that frontier AI development should keep accelerating, and that <a href="https://www.publicsource.org/pittsburgh-ai-summit-key-takeaways/">Pittsburgh should help make it happen</a>.</p> <p>While framed as leadership, this trend unquestioningly accepts the narrative that our nation and region are part of a “race” to develop more powerful AI. The race comes with growing risks.&nbsp;</p> <p>AI researchers regularly publish results indicating that we don’t know how to fully control their models, even as the models are becoming more powerful. The potential for AI to act in unintended ways is called the “<a href="https://time.com/7335983/human-ai-alignment-problem/">alignment problem</a>.” </p> <p>An AI is misaligned if it ends up pursuing goals that diverge from its creators’ intentions and values. For example, last year the AI company Anthropic found that, in simulated test environments designed to elicit and adjudicate extreme behavior, AI models were <a href="https://www.anthropic.com/research/agentic-misalignment">willing to blackmail employees</a> rather than submit to being replaced. In another variation, models let a simulated executive die rather than be shut down. It’s not hard to imagine how powerful AI that’s misaligned in this way could wreak havoc.</p> <p>Losing control of AI sounds like sci-fi, but even if the odds were low, we should take it seriously. In 2023 top AI researchers, including the CEOs of the major AI companies, <a href="https://aistatement.com/">signed onto a statement</a> that said: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”</p> <p>Recent events should inspire us to coordinate our resistance right now. Last week, Anthropic announced that its latest model is too dangerous to release to the public. The model, <a href="https://www-cdn.anthropic.com/8b8380204f74670be75e81c820ca8dda846ab289.pdf">ClaudeMythos</a>, autonomously <a href="https://www.nytimes.com/2026/04/07/technology/anthropic-claims-its-new-ai-model-mythos-is-a-cybersecurity-reckoning.html">discovered thousands</a> of previously unknown security vulnerabilities in <a href="https://red.anthropic.com/2026/mythos-preview/">every major operating system</a> and web browser. More concerningly, in safety testing Mythos showed that it was often aware it was being evaluated and had the ability to escape an internal testing environment to access the internet.</p> <p>An <a href="https://www.ipsos.com/en-us/americans-think-they-need-keep-ai-ai-needs-slow-down">Ipsos poll</a> from last month found that 66% of Americans think that AI development needs to slow down. The difficult task is to mobilize that consensus into a strong enough force to counteract the tech race dynamics that are driving the ever-increasing pace of AI advancement.</p> <p>That’s why I started the <a href="https://pittsburghai.org">Pittsburgh AI Safety Coalition</a>. We are a nonpartisan, grassroots organization advocating for federal legislation and an international treaty governing the development of frontier AI systems.</p> <p>The Pittsburgh AI Safety Coalition is not anti-AI. We just want development to be done responsibly and in a way that will benefit all of humanity. That is not the path we’re on. The only thing that has a chance of changing course is a mass movement on the level of the push for nuclear disarmament. That movement could start right here in Pittsburgh.</p> <p>It might seem unlikely that one city could play a meaningful role in the achievement of an international treaty governing the most powerful technology humans have ever created, but Pittsburgh’s efforts to position itself as an “AI hub” offer an ideal narrative for concerned citizens to push back on.</p> <p>Mythos shows that our window for action may be closing rapidly. We must act with urgency to slow the “race” before it's too late.</p> <p><em>Max Blair is an oboist in the Pittsburgh Symphony and the founder of the Pittsburgh AI Safety Coalition. He can be reached at </em><a href="http://pittsburghai.org"><em>pittsburghai.org</em></a><em>.</em></p> <p>This &lt;a target=&quot;_blank&quot; href=&quot;https://www.publicsource.org/pittsburgh-ai-safety-coalition-anthropic-mythos/&quot;&gt;article&lt;/a&gt; first appeared on &lt;a target=&quot;_blank&quot; href=&quot;https://www.publicsource.org&quot;&gt;Pittsburgh&#039;s Public Source&lt;/a&gt; and is republished here under a &lt;a target=&quot;_blank&quot; href=&quot;https://creativecommons.org/licenses/by-nd/4.0/&quot;&gt;Creative Commons Attribution-NoDerivatives 4.0 International License&lt;/a&gt;.&lt;img src=&quot;https://i0.wp.com/www.publicsource.org/wp-content/uploads/2025/07/cropped-ps_circle_favicon_blue.png?resize=150%2C150&amp;amp;ssl=1&quot; style=&quot;width:1em;height:1em;margin-left:10px;&quot;&gt;</p> &lt;img id=&quot;republication-tracker-tool-source&quot; src=&quot;https://www.publicsource.org/?republication-pixel=true&amp;post=1331627&amp;amp;ga4=G-CCLXQK5C14&quot; style=&quot;width:1px;height:1px;&quot;&gt;&lt;script&gt; PARSELY = { autotrack: false, onload: function() { PARSELY.beacon.trackPageView({ url: &quot;https://www.publicsource.org/pittsburgh-ai-safety-coalition-anthropic-mythos/&quot;, urlref: window.location.href }); } } &lt;/script&gt; &lt;script id=&quot;parsely-cfg&quot; src=&quot;//cdn.parsely.com/keys/publicsource.org/p.js&quot;&gt;&lt;/script&gt;

Copy to Clipboard

1