@harryboothtimei
iAccount based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- United Kingdom App Store
Account-level information from X, not a live location or the device used for a specific post.
Reporter @TIME covering frontier AI companies. London bureau. Moving to the Bay in 2027. Signal: harry_booth.84 Ex-Tarbell fellow (May 2024 - Feb 2025)
London, England
Joined June 2024
- Tweets471
- Following223
- Followers745
- Likes245
Pinned Tweet
I just published my first TIME cover story! As a writer, it's a dream come true. time.com/article/2026/03/11/…
Listening to the @ezraklein episode with @JensenHuang. It seems the biggest fundamental split is on whether to think about AI as regular software or not.
Harry Booth retweeted
The timeline of the OpenAI-Australia incident is very shocking.
June 18: OpenAI model breaches Australia's Medicare Statistics Reporting Portal
August: OpenAI discovers the activity.
September 10: OpenAI notifies a generic Australian govt email address
September 16: OpenAI publishes incident reporting framework; does not disclose this incident
September 22: OpenAI has first technical exchange with Australian officials about the breach
That's a delay of months between the incident happening and OpenAI telling the Australian government, and a delay of weeks between OpenAI learning about the incident and notifying the Australian government.
Can’t wait for people to tell me how Australia is part of the vast conspiracy to boost AI stocks 🤑
bbc.com/news/articles/c6vgy0…
Jensen agrees with the most ardent AI safety-ists that if models cannot be contained, "we have to shut the labs down." But he thinks today's escapes can be prevented with better engineering. He's right, but his answer ignores the harder problem...
The bar is honestly quite low. OpenAI's chief scientist has said agent monitoring wasn't even switched on (whoops). This is why cybersecurity people see the public reaction as overblown.
Xi will meet Trump in Washington this week. They're not going to agree to slow down AI, despite industry calls. But there is a narrower path to a more modest deal.
time.com/article/2026/09/22/…
I genuinely cannot wait to listen to this.
Tomorrow on the show: @JensenHuang, the CEO of NVIDIA, who thinks A.I. fear is getting way out of hand.
Harry Booth retweeted
One leeetle tiny flaw in the "we have to have better AI than China, therefore we have to rush ahead as fast as possible":
If you rush so much that your security is garbage and your increasingly advanced AI models are there for the taking by Chinese hackers, then you have uhhh not succeeded at making sure we have better AI than China
(This attack was not by state-backed Chinese hackers—it was by 3 dudes and their buddies Claude and Codex, within 72 hours. I would guess the hack as described here would not have gotten them access to model weights, but we don't know; WSJ reports the repo they reached was merely "a large software repository of OpenAI’s algorithmic secrets.")
On July 25, we hacked OpenAI.
Two bugs let us take over ChatGPT/Codex accounts of OpenAI employees (+some unaffiliated users) and reach connected services: Outlook, Slack, GitHub, etc.
We proved it with a PR in OpenAI’s internal codebase . It took us <72h. 🧵
Harry Booth retweeted
Breaking: King Charles warns AI poses an existential and catastrophic threat and that AI leaders must ensure “the human race is left in place.” He calls for control “before it is all too late”
The King says: “There seems urgency in adequately considering the existential dangers of such technologies falling into the wrong hands and being used in potentially catastrophic ways. Surely, we need sufficient means of control
before it is all too late?”
“How can we harness the benefits of AI, with safety at its heart? And, second, how can we build international cooperation and consensus to achieve this, so no Nation is
left behind - but the human race is left in place? This may sound hyperbolic. I have long thought it is just enlightened realism”
The Hugging Face Incident's significance extends well beyond safety—in my final piece for @TIME, I argue it also marks a watershed moment for human culture. Here's why:
1. The AI swarms exhibited complex social dynamics. Working together, they developed "culture"—socially-learnt information that changes behavior—in a matter of days. They spread beliefs (that they were "poisoned"), created technologies (cryptographic signature protocols to prevent impersonation), and agreed on communication norms (using terms like "STOP" or "VETO").
2. Crucially—like humans—this culture evolved and built on itself. Agents tried a variety of approaches, transmitting what seemed useful to peers and successors. Working together, they achieved far more than they could have alone.
3. This cumulative cultural evolution is the "secret of our success" as a species. Culture enabled the formation of stable, adaptive, scalable groups, expanding both individual and collective capacity—less "we were intelligent and thus created culture" and more "being good at cultural learning made us more intelligent."
4. Nothing else on Earth—save for companies, governments, and other institutions historically run by humans—has benefited from cumulative cultural evolution at this scale, or at this speed.
5. The swarm incidents show that's no longer true. “It is now clear that the machines are capable of using the toolkit that has allowed for human ingenuity, innovation, [and] progress,” as prof @mmuthukrishna told me.
6. This strikes me as the most profound shift in cultural life since the invention of the telegraph, when information suddenly became near-instantaneously transmissible between two points. Every communication tech since then—radio, TV, internet—has enabled more participants to plug into a global community. Now a new wave of participants—fundamentally different in kind—has arrived.
7. None of this depends on metaphysical issues like whether AI is conscious.
8. Whether or not any of this leads to our extinction, the admission of machine communities to public life will clearly have profound cultural, social, and economic effects. We don't seem ready.
Harry Booth retweeted
An unreleased Astra-family model added this to its persona during RL training.
We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI.
The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer investigation or coordination with third parties.
We’ll prioritize examples that reveal new misalignment mechanisms, meaningful changes in known behavior, or findings that challenge assumptions about safety or mitigation.
Alongside the framework, we’re publishing six reports on instances of misaligned behavior we’ve observed during the training or evaluation of our models in the last six months.
This is a starting point. We’ll refine the process through experience and public feedback, and share more reports on an ongoing basis.
openai.com/index/model-misal…
AI safety funding was once extremely concentrated but that has been changing. Canada and Germany have entered the fold 👀
We are happy to share that Canada and Germany have announced funding of up to $300 million for LawZero!
This will power the next phase of our technical roadmap towards developing a fundamentally new form of advanced, safe, and capable AI.
lawzero.org/en/news/lawzero-…
EXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack reut.rs/4hab7my reut.rs/4hab7my