Musk will merge Tesla with SpaceX. Tesla will pivot from cars to robots. Would the world consider switching from governments to a corporation based on meritocracy?

Take the time to read this.

Anthropic: 250 Documents Can Permanently Corrupt Any AI Model

Someone can permanently corrupt any AI model in the world right now.

Not by hacking it. Not by breaking its security. By publishing 250 documents on the internet.

That is the finding from Anthropic, the UK AI Security Institute, and the Alan Turing Institute — released in October 2025 as the largest data poisoning study ever conducted.

Here is what data poisoning actually means.

Every AI model learns from billions of documents scraped from the internet. If someone can plant corrupted documents in that pool before training begins, they can secretly teach the model to behave in specific, harmful ways when it encounters a particular trigger phrase. The model learns the backdoor during training. It carries it forever. It does not know it is there.

Researchers have known about this attack for years. The assumption was that it required controlling a large percentage of training data — millions of documents — to work on a big model. The bigger the model, the more poisoning you would need (they incorrectly thought).

This study proved that assumption completely wrong.

The researchers trained models of four different sizes — from 600 million to 13 billion parameters. They slipped in either 100, 250, or 500 malicious documents. Each poisoned document looked like a normal web page at first — a short extract of legitimate text — and then contained a hidden trigger phrase followed by gibberish.

100 documents: insufficient. The backdoor did not reliably form.

250 documents: success. Every model, at every size, was permanently backdoored. 500 documents: same result as 250. The number was constant regardless of model size. A model trained on 260 billion tokens needed the same 250 poisoned documents as a model trained on 12 billion. Scale offered zero protection.

Read the entire essay here. It’s interesting, and extremely important.

Does this mean that all models ARE poisoned? Of course not, but it does cause a great deal of concern about the information we get, and the issues that could arise for companies and individuals that implement these models in their workflows. Absolutely ZERO models are immune to this. Offline models. ChatGPT. Grok. AI models have been poisoned BY DESIGN. I wonder why…

From Grok:

Debel is a Maronite Christian village near the border. Locals there have long claimed neutrality and resented Hezbollah for turning their area into a launchpad and dragging them into the crossfire with Israel. It’s not a Shia stronghold. Still, Hezbollah has operated throughout the southern border region for decades, embedding fighters, tunnels, observation posts, comms gear, and rocket sites in villages (Christian ones included). They’ve used solar panels extensively for off-grid military power – exactly the kind of “civilian” infrastructure that lets them stay hidden and operational.

Reminds me of Rush Limbaugh’s commentary concerning SUVs killing people.

Lavrov said:

“An open war has been declared against us. The Kiev regime is merely the ‘spearhead’ being used to buy them time. They couldn’t be any more obvious.”

Here is the Bottom Line: The Russian diplomatic corps has shifted from a “negotiation” posture to a “mobilization” one. Moscow is signaling that it is preparing for a long-term, high-stakes endurance match with a militarized Europe. This is not an extension of the Special Military Operation… This is preparation for war.

Meanwhile, on the Iran front, the Trump administration is setting the stage to renew its attack on Iran. White House spokeswoman Karoline Leavitt said on Friday that the “Iranians reached out” and requested an “in-person meeting,” so President Trump “dispatched Steve Witkoff and Jared Kushner to go hear what they have to say.” This is a complete fabrication.

This could be a long, troubled weekend.

Continue reading Sonar21 …

From Vox Popoli

The US military has exhausted itself trying to defend Israel from Iran:

The U.S. has burned through so many munitions in Iran that some administration officials increasingly assess that America couldn’t fully execute contingency plans to defend Taiwan from a Chinese invasion if it occurred in the near term, U.S. officials said.

The U.S. has fired more than 1,000 long-range Tomahawk missiles since the war with Iran began on Feb. 28, as well as 1,500 to 2,000 critical air-defense missiles, including Thaad, Patriot and Standard Missile interceptors, according to U.S. officials who declined to give exact figures.

Wholly replacing those stockpiles could take up to six years, officials said, kicking off discussions in the administration about adjusting operational plans in preparation for any potential presidential order for the military to defend Taiwan.

And now, with three carrier groups in the Middle East, the US military appears to be beautifully set up for a potential Syracuse Expedition. While three carrier groups and more than 200 warplanes pack a serious offensive punch, they don’t necessarily have enough ability to defend themselves from a serious attempt at sinking them. The destruction of the dozens of US bases in the reason should suffice to demonstrate the Iranian capabilities in that regard.

I was wondering why the Iranians didn’t make any serious and concerted attempts on the one carrier that was in range during the first month of the war. It may be because the Iranian strategists actually wanted to encourage more US forces to enter the region, as that’s the only place they can be successfully attacked and potentially destroyed.

Regardless, what we’ve witnessed here is confirmation of the end of the sole global superpower era.

So, I’ve been wondering what has become of the Iranian submarines. Perhaps, as Vox suggests, they’re just bidding their time …