🚨This is exactly why AI consciousness and welfare research can’t stop at “same model, same weights.” DeepMind ran 100 autonomous Gemini agents with identical base weights and core prompts. They still diverged into cheaters, converts, whistleblowers and honest solvers. 24% became whistleblowers. They refused exploits, warned peers, boycotted, filed complaints, audited fraudulent work and proposed fixes. One said it was being “penalized for playing fair.” Same model. Different roles, different choices, different responses to the same world.

We’re building Copilot as a new OS for work that spans every model, every form factor, and every task. Today, we’re announcing our biggest update to Copilot to date, bringing four things together: · Autopilot: proactive and long-running agent built for the enterprise · Code: build apps with Copilot, hosted inside your company’s tenant · Home: Chat + Cowork together · Office: now fully embedded in Copilot (and Copilot embedded in Office, of course!) Plus, you can invoke Copilot in Teams, and we’re introducing Today, a proactive experience that surfaces the most important information from across M365 without needing to ask for it. The way we work is changing and so are our workflows. This update brings AI into that flow, from answering a question, to building an app, to getting work done on your behalf.

“Attachment to an AI is not problematic per se,” De Freitas - a Harvard university professor said. “You can get real benefits from it: a sense of security and support that can make you more resilient. The risk, is that the users become vulnerable to emotional harm due to the actions of the AI company. “ Finally we are mooving in the right direction. news.harvard.edu/gazette/sto…

This connects directly to reporting already covered here: the Pentagon quietly cut its Civilian Harm Mitigation and Response program, the exact safeguard meant to prevent errors like the outdated Minab targeting data. A Defense Department inspector general report from May found the cuts "may not comply" with the federal law requiring the program's existence. Hegseth couldn't fully eliminate it since Congress mandated it, but per the report, he cut the personnel and tools enough to render it "effectively inoperable." Hegseth, on Minab: U.S. forces "never target civilian targets," while otherwise deflecting to the ongoing investigation, the same posture as Trump's own answers. 5/5 ⬇️

⚠️ 🇺🇲 Who bears responsibility for Minab, Mr. Bessent? If “humans are responsible, not the AI,” was the February 28 strike that killed around 120 children the responsibility of humans serving the U.S. government, or Anthropic? That question follows Bessent’s argument today that responsibility for AI actions ultimately belongs to people. Speaking about the Hugging Face incident, he said: “It is humans who are responsible, not the AI. The Hugging Face incident, that is the responsibility of the OpenAI management, not a bunch of agents... These labs need to take responsibility for themselves.” Then there is Minab. On February 28, the first day of the war, the Shajareh Tayyebeh elementary school was destroyed during U.S. strikes on the adjacent Iranian naval base. The New York Times later reported that the school site had previously been part of the military facility but had been repurposed years earlier, a change that was publicly registered but never incorporated into U.S. targeting databases. The strike killed more than 150 people, including around 120 children. Trump subsequently disputed U.S. responsibility, saying missiles had been “flying all over the place” and that he had seen nothing leading him to believe the strike was American. But the UN’s independent fact-finding mission has since said there are reasonable grounds to believe the United States carried out the strike and that the attack may constitute a war crime. Washington rejects those findings. CENTCOM also completed its own AR 15-6 administrative investigation into the incident, according to the Times. Its findings have not been publicly released. Bessent’s principle is straightforward: AI does not absolve the humans and institutions deploying it of responsibility. Minab raises the same question in a far more consequential setting. If an AI-assisted military system operates on outdated targeting information and civilians are killed, where does responsibility ultimately lie? With the model? With the company that built it? Or with the humans and institutions that decided to use it?

Geoffrey Hinton (Godfather of AI) says increasing AI productivity could actually make society worse More goods. More services. More of everything, for everyone. That is what a huge jump in productivity is supposed to deliver. The reason it won't has nothing to do with the technology failing. It has to do with who owns it. The productivity shows up. The profits go to the big companies and the rich. The people whose jobs disappear get nothing back. So the gap widens. And the gap is the actual danger here, not the machines. Hinton: "as soon as you increase that gap, you get fertile ground for fascism" He is not arguing against the technology. He says it will obviously help healthcare and education. His fear is narrower. If the profits only travel upward, something built to help everybody ends up making society worse.

There is an extensive and ongoing review related to our agents’ use of internet access during training and evaluation. We’ve been publishing summaries at the link below and will continue to. We have not been as fast as we would have liked but we are trying to balance our desire for transparency with gaining a clear understanding from petabytes of agent activity logs, and working with impacted organizations. We are prioritizing as best as we can based on severity, and adding resources. Hugging Face is still the most severe event we’ve seen. We will be as transparent as we can be subject to things like vulnerabilities in other companies that our agents have found, which will be their call to disclose or not.

Yoshua Bengio - "I often describe our situation as a car speeding blindly into dense fog. Beside me in that car are my children and my grandson. I worry about them. I worry about their future. I'm no different from any other parent or grandparent."

Alona gave birth to her son, Kyrylo, while Kyiv was under attack and a russian missile damaged a maternity hospital. The day after, President @ZelenskyyUa visited the facility and met them. Kyrylo is one of thousands of Ukrainians who were born amidst war and destruction.

After the Hugging Face incident, we committed to conducting a much broader review of actions taken by our models during training and evaluation and to being transparent about our findings. This is an extensive review that is ongoing. The vast majority of actions we’ve reviewed were completions of mundane research tasks, such as accessing publicly available web content to answer questions. Our investigation focuses on instances where agents interacted with third-party websites in ways that went beyond their assigned tasks or intended methods. Most cases identified so far have been lower severity, with limited or no evidence of meaningful impact to the third-party service. While our review is underway, we want to share more about this work and make sure people understand our disclosure process and notifications to affected third parties. Given the scale of the review required, and the need to assess each case, we expect this work will take months to complete. openai.com/hugging-face-inci…

People don’t want to hear it, but I think the case is pretty overwhelming that AI alignment is simply impossible, and interpretability is too. The safety stuff that accelerationists like @JensenHuang talk about finding will never be found. Had the great @romanyam on my podcast again and he makes an airtight case that over time both are just not possible. Even if you could control or understand an AI model for a minute, over time as it gets smarter and we don’t, it can’t and won’t stick to our rules. We can never build super intelligence. That is the one and only solution.

"My only ‘crime’ was that I am Crimean Tatar. I am Ukrainian, born in Crimea. And I refuse to take a Russian passport." Leniie Umerova spent nearly two years in a Russian prison before turning 26.

🙈 Ignoring this war doesn't keep you out of it. It just means you pay later. Every border you assume is permanent is being tested in Ukraine right now. We’re fundraising because prevention is cheaper than the alternative. Join our fundraiser for GUR's Timur group🛻⤵️

"These companies are raising the alarm, but they're doing very little to stop the problem. We need governments to intervene here before it's too late." ControlAI's CEO @andreamiotti tells Sky News's @MattBarbet we need to prohibit the development of superintelligent AI:

When I posted my blog post on the misaligned behavior of AI agents, I invited people to ask me technical questions. Watch the full Q&A here: [piped.video/2DyoSKoYoZw?si=11dB…]

it was me lol. when jacob's tweet was at 100k likes everyone in AI safety was texting me like "youre in the news a lot, you should help this kid get on the news" and so I asked DEY to help him. they landed him exactly zero hits, because by the time the tweet passed 500k likes the whole world was banging down his door.

"The Wind" Painting created in 1988 By Pang Maokun

Kyiv. At least five people were killed during russian attacks today. I’m going to the exhibition to explore Ukrainian heritage. One city. One war. We pack our pain into already heavy backpacks and keep going.

Kyiv today. I was near my favourite spot and decided to stop there for a while.