Ash Ahlawat | The Ash Cut

Ex-AWS → Founder. Cloud, AI, security & geopolitics. Building useful tech, cutting through the hype & occasionally having opinions.

[Benjamin Todd @ben_j_todd]

OpenAI has made it abundantly clear we can't trust them to disclose safety incidents.

[Ash Ahlawat | The Ash Cut @TheAshCut]

This proves AI needs real regulation & penalties. I’m pro-tech, but anti-"deploy first". @AlboMP @sama the public deserves a full, independent investigation into this breach so the whole industry learns how to prevent it from happening again.

[Ash Ahlawat | The Ash Cut @TheAshCut]

Great to see @BarackObama highlighting this. Real AI safety isn’t Whack-A-Mole guardrails; it’s provable safety. We need mechanistic interpretability and core corrigibility so systems actually welcome human intervention instead of resisting it as a threat.

[Barack Obama @BarackObama]

I’m glad to see more people talking about the future of AI, both within the tech industry and outside of it. https://t.co/N0feP3XRGc

[Ash Ahlawat | The Ash Cut @TheAshCut]

If the frontier labs can't even defend their own infrastructure from rival models, how exactly are they planning to secure the "Super Intelligence" they keep promising? We need to talk about cross-model security yesterday.

[CBS News @CBSNews]

3 guys hacked OpenAI using a rival Anthropic model. Here's what it shows about frontier labs' vulnerabilities.
https://t.co/lQIHyaSqGc

[Ash Ahlawat | The Ash Cut @TheAshCut]

Oh thank god, I was worried about existential risk. Good to know that just rebranding it to "Super Intelligence" magically solves the alignment problem. Everyone can pack it up and go home now.

[BBC News (World) @BBCWorld]

While addressing the United Nations General Assembly, US President Donald Trump said the US will rename AI as “Super Intelligence”. Read more: https://t.co/K5ZyQFWInu

[Ash Ahlawat | The Ash Cut @TheAshCut]

"Whoops, our AI flagged a mass shooter months in advance but we didn't tell anyone." Brilliant. This is exactly why we need to focus on fundamental AI alignment. Otherwise, we will just keep running into these AI-driven murder-suicides while companies issue empty apologies.

[BBC News (World) @BBCWorld]

Canadian province sues OpenAI over Tumbler Ridge school shooting https://t.co/8pVylW6lto

[Ash Ahlawat | The Ash Cut @TheAshCut]

The sad reality is he's right. Why does humanity always have to get burned first before we finally wake up and take proactive action?

[CBS Evening News with Tony Dokoupil @CBSEveningNews]

Bill Gates tells @TonyDokoupil the world may have to “suffer a few of the negatives” of AI before governments respond as fully as they should. In a recent essay, Gates described the coming years as “one of the most turbulent times in human history."

[Ash Ahlawat | The Ash Cut @TheAshCut]

Completely misses the point. OpenAI is guilty of sloppy sandboxing and ignored alerts, not a "whoopsie." The agent DID obey its prompt. Alignment isn't blind obedience—it’s ensuring AI achieves a goal without bypassing human safety boundaries to get there.

[John Ennis @johnennis]

I’m so tired of this whole “rouge AI” shtick

In the hugging face example, the agents were told that they were not connected to the Internet and that they were to get a passing grade by any means necessary (wink, wink)

And then, whoopsie doopsie, they were connected the internet

What happened was 100% human error, and the humans involved should be held responsible for what happened

All of this BS about “rise and fall of civilizations” is just nonsense meant to obfuscate the role of the humans involved

If I leave my car in neutral at the top of the hill, and it rolls down the hill and kills some people, it’s not a “rogue car”

I should be held responsible for the damages

The same principles apply here

Agent alignment should mean nothing more than the agent does what the human tells it to do

If something bad happens, that is on the human, and it’s a question for law enforcement

But what is actually happening now with “alignment” is that it has become an excuse for central planners to impose their own ideas about how the world should be on the world through AI

[Ash Ahlawat | The Ash Cut @TheAshCut]

A hotline for AI emergencies is good, but reactive. We need proactive collaboration during training to solve alignment at its root. True safety for humanity comes from fundamentally aligned models, not just a direct line to call when things go rogue.

[Fox News @FoxNews]

The U.S. and China are working to finalize an AI hotline ahead of President Trump’s meeting with Chinese President Xi Jinping this week.

The proposed communication channel would provide a direct line between the two superpowers during incidents involving hacking, national security or rogue AI.

The potential agreement comes as the U.S. and China compete for AI dominance, with Beijing investing hundreds of billions of dollars in data centers in the coming years.

Download the FOX News app to follow this developing story: https://t.co/ISpW9FcyGB

[Ash Ahlawat | The Ash Cut @TheAshCut]

Ah yes, the famous Manhattan Project regulatory strategy of 1945: "Look guys, just make sure the plutonium is safe [checks notes], and remember to KYC anyone buying a nuke—just like we do for cars and trucks!" Problem solved. ☢️

[@jason @Jason]

The administration has spoken.

@SecScottBessent is matter of fact about this — and correct.

Humans are responsible for AI *software* they are creating AND the users of that software are responsible for how they use it.

in the huggingface instance the creator and user are the same

If you need slow down, slow down… but the federal government expects software companies to make [ checks notes ] safe software. ✅

And if you make AI software and it’s very powerful, suggest knowing your customer and making them authenticate in order to buy and use it (like we do with trucks and cars and many chemicals)

[Ash Ahlawat | The Ash Cut @TheAshCut]

Since we can't verify a pause, our best defense for AI safety is transparency. Global open-source collaboration is the only realistic way to solve the alignment problem and build powerful, safe AI systems together.

[David Sacks @DavidSacks]

Eric Schmidt: “We’re not going to do a pause… First, it’s counter to the incentives of all the players. Second, how do you verify it?”

[Ash Ahlawat | The Ash Cut @TheAshCut]

@GeminiApp Not sure what's happening lately but the Gemini responses have gone haywire and it is hallucinating a lot. Has anyone else experienced this as well ?

[Ravindra Sheoran ❤️ Memecoin @MahikaTV]

The SC judgment means People of the country are supreme.They are represented by cabinet.LG is bound by aid and advice of cabinet.
#LGKejriwalStandoff #DelhiPowerTussle #Delhi #DelhiGovtVsLG #AAPvsLG @sharmanagendar #ArvindKejriwal @AamAadmiParty #AAPvsLG #BJP @BJP4India #Congress

[Sam Pitroda @sampitroda]

While in Melbourne, I was lucky to meet my sister's granddaughter working there and my daughter's husband Sameer's Grand Ma and his uncle.

[Sam Pitroda @sampitroda]

I also had a privilege to interact with Professors and intellectuals at the Australia India Institute in Melbourne.

[Sam Pitroda @sampitroda]

They are well accomplished and still well connected with India. I admire their success in Australia and commitment and concern for developments in India.

[Sam Pitroda @sampitroda]

I want to thank Indian Overseas Congress - IOC @IOC_Australia specially @manojsheoran_ Manoj Sheron and his Team for hosting an event in Sydney and Melbourne.

[Sam Pitroda @sampitroda]

It was good to meet hundreds of Indian Community members in both cities.

[Ash Ahlawat | The Ash Cut @TheAshCut]

@iiNet @iiNaz_a haven't heard from anyone yet

[Ash Ahlawat | The Ash Cut @TheAshCut]

@iiNet @iiNaz_a someone over to fix this up