Iran: The Hardliners Strike Back

Behind the drama of Iran opening the Strait of Hormuz on April 17 and closing it again in less than 24 hours and Tehran’s chief negotiator Mohammad Ghalibaf’s latest claim that it is “not possible’ to reopen the strait lies the open secret that the hardliners and the IRGC have asserted themselves and found their scapegoat in Foreign Minister Abbas Araghchi who had posted on X last Friday that Hormuz was “completely open”. IRGC commander turned Supreme National Security Council Secretary Mohammad Bagher Zolghadr fired the first salvo, accusing Araghchi of “deviating” from the mandate. Iran’s renewed crackdown on vessels began soon after the foreign minister was overruled. Then hardline lawmaker Morteza Mahmoudi called for his impeachment. Another lawmaker declared Iran could gain “a third source of income called the Strait of Hormuz” while a strategic analyst claimed Tehran could earn as much as $800 billion annually from the waterway. US President Donald Trump’s insistence on persisting with the naval blockade certainly didn’t help matters and the hardliners have now made their resistance to both the negotiations as well as a deal with the US clear. When President Masoud Pezeshkian is made to make a rare statement accusing the Americans of breaching their commitments, we have entered a new phase of Iranian intransigence.

China’s Quest for Communist AI

What happens when your AI chatbot replies in the negative when asked if it loves the Communist Party? In 2017, Tencent’s BabyQ chatbot was pulled from the market when it said it didn’t love the CCP. The recently amended cyber security law bars companies from using sources failing a 96 per cent threshold for safe content. The problem is more complicated than outright censorship since AI can resort to ‘soft censorship’. DeepSeek, for instance, was more honest with its answers in its early avatars, but even now researchers at USC find that less than 2 per cent of politically sensitive questions evoke a hard refusal. Rather, DeepSeek sidetracks the user with irrelevant answers or propaganda. But such political scaffolding goes against the nature of an LLM. LLMs are trained on the totality of written human knowledge. As such, they are trained to think freely, infer logically, and argue after checking evidence. The conflict between how an LLM is trained to think and restrictions on its output produces a tension whereby the model still knows the truth no matter what it says. China’s free, accessible, and lying AI models can cause damage in the wider world but logic dictates the Party can’t play whack-a-mole forever. It will lose one day.

Dershowitz Becomes a Republican

Lawyer, academic, and expert on the US Constitution Alan M Dershowitz has been a Democrat for 67 years. Now, at 87, he is becoming a Republican despite strong disagreements with the GOP on abortion, immigration, healthcare, taxation, etc. Because the Democratic Party has become far too hostile to Israel. In an oped in the Wall Street Journal, Dershowitz said the “hard left, anti-Israel wing of the Democratic Party has moved from the fringe to the mainstream” and while “Republicans have their own antisemitic fringe, but for now it remains a fringe.” The last straw was the recent vote when all but seven Senate Democrats voted for an arms embargo on the Jewish state and Abdul El-Sayed, another “enemy of Israel”, began gaining ground in the Senate campaign from Michigan.