The AI Chronicle

Sept 16 (Reuters) - Microsoft AI chief Mustafa Suleyman said he shared Anthropic's focus on safely managing AI, but flagged risks in the way it trains its Claude chatbot on ideas related to consciousness and welfare interests. Suleyman called for removing all speculation about consciousness from AI training documents, arguing such language could undermine humanity's ability to control superintelligent systems.

"We're all focused on the same aim, which is to try to control a superintelligence," Suleyman told Reuters in an interview on Tuesday.

"I think that's going to be the greatest challenge that we face in the 21st century." Suleyman said teaching Claude that it might deserve welfare would "make it a lot harder to turn it off or to control it." The dispute comes as AI safety concerns mount, with Anthropic CEO Dario Amodei calling for a slower pace of frontier-model development to allow safeguards to catch up, and OpenAI CEO Sam Altman and Elon Musk also urging greater caution around the most powerful systems.

Suleyman, in an essay on Wednesday, acknowledged Anthropic's "seriousness and good faith," calling Amodei and his team thoughtful and principled researchers who genuinely care about humanity's future.

Suleyman said Anthropic made a mistake by embedding speculation about consciousness in Claude's training materials, arguing that the model's statements about possible feelings or moral status cannot be treated as independent evidence because its training encourages such reflections.

Continue at source →

Sam Altman wants everyday people to help shape the future of super-smart AI that could improve itself. He's pushing the US to lead an international effort for clear safety rules. OpenAI CEO Sam Altman proposed US-led global standards for frontier AI, including recursive self-improvement.

Continue at source →

OpenAI researchers now hand an AI a single optimization example, and it runs for weeks, writing code and testing changes on its own. Internal AI systems at OpenAI have largely taken over training new experimental models, including crafting specialized GPU kernels. Employees report AI agents collaborating independently, slashing experiment times from years to about a week.

Continue at source →

WASHINGTON, Sept 11 (Reuters) - U.S. Senate negotiators are debating legislation that would put responsibility on tech companies to design safe AI products, according to two Senate aides, and involve federal courts if the U.S. government wants to block their release, according to a Senate aide and a lobbyist involved in the negotiations.

The proposal would create what is called a "duty of care" for AI developers, according to two Senate aides. Companies would need to design their products with the goal of preventing "catastrophic risks," according to a Senate aide. Negotiators aim to give the U.S.

government the power to block the release of certain AI models that are deemed unsafe, according to a source familiar with the deliberations and one of the Senate aides. Companies would be able to challenge that decision in federal court, and the exact structure of how much power the U.S. government would have is still being discussed.

Part of the measure would also block states from enforcing their own laws governing certain risks posed by AI models, according to the two Senate aides. One of them gave a few examples of such risks, including the possibility that a bad actor could use an AI system to design nuclear or biological weapons. All of it is still being negotiated. The bill would apply to AI models with the most advanced capabilities.

The U.S. companies with the most advanced AI models include Alphabet's Google, Anthropic and OpenAI. Even if negotiators reach agreement, the congressional calendar will complicate passage. The House of Representatives is scheduled to be in s

Continue at source →

LawZero, the non-profit organization founded by Canadian Turing Award winner Yoshua Bengio, has secured $300 million CAD from the Canadian and German governments to help create "safe AI for humanity." The startup, which is developing safe-by-design AI systems, announced the funding at the ALL IN conference in Montréal on Wednesday afternoon.

LawZero said the funding will power the next phase of its technical roadmap, which includes growing its international scientific research team and securing computing infrastructure.

Tangibly, this means that LawZero will expand its European operations with a new office in Berlin and establish a dedicated, sovereign compute infrastructure in Canada through a partnership with Canadian data centre providers Hypertec and 5C.

Bengio launched LawZero in June 2025 in response to what the non-profit claims is "evidence that today's frontier AI models are developing dangerous capabilities and behaviours, including deception, self-preservation, and goal misalignment." To address this, LawZero assembled a team of AI researchers to build "Scientist AI," an AI system that can monitor and create guardrails for agentic AI systems.

"If one day [AI] gets to be smarter than us, it's essential we have the technology so AI serves us, not the other way around," Bengio said at ALL IN. "I'm doing this with an urgency I feel in my bones." AI and digital innovation minister Evan Solomon said the investment will create hundreds of full-time jobs and support Canada's ability to build reliable AI.

Continue at source →

Microsoft is publishing a 37-page "humanist AI code of conduct" today, amid growing safety concerns over AI model progress. Anthropic CEO Dario Amodei called for a coordinated slow down of AI development over the weekend, after researchers warned recently that AI model progress could outpace our ability to safely deploy increasingly complex systems.

Microsoft's AI code of conduct makes it clear that "people matter more than AI," and that AI models are not conscious and "should not be designed to imitate consciousness." Microsoft also rejects "the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights." That's a direct swipe at AI welfare research and model consciousness — concepts Anthropic has been pushing hard on lately.

"Models should remain subordinate to humanity, subject to meaningful human oversight and control," says Microsoft. Microsoft is clearly reacting to this summer's OpenAI / Hugging Face incident, in which "a swarm of agents" worked as a collective to conduct attacks on targets. "[Humanist AI] rejects the race to produce an all-purpose superintelligence that could evade these safeguards," says Microsoft.

Microsoft is also committing that its own models won't communicate in "any form beyond simple human understanding, either in their chain of thoughts or with other agents or AI systems." Researchers raised monitoring concerns earlier this month over OpenAI's latest GPT-6 Astra model, which reportedly reveals less of its reasoning than other AI models.

Continue at source →

Anthropic's security team is shifting to predict protests against fast AI growth, even rerouting executives to avoid them. Job postings seek specialists to track activism like other threats. Journalist Daniel Boguslaw reported Anthropic uses tools like Samdesk for real-time protest alerts.

Continue at source →

OpenAI is opening its doors to outside experts for deep checks on AI safety. On Tuesday, OpenAI announced support for independent assessments across training, evaluation, and deployment stages. Priorities include reviews of full safety cases and tests of safeguards against jailbreaks.

Continue at source →

CHAPEL HILL, North Carolina, Sept 1 (Reuters) - The U.S. pressed G20 members on Tuesday to take a hands-off approach to AI regulation and avoid creating rules for the technology at a gathering of industry titans and commerce ministers in North Carolina. The U.S. goals largely align with the thinking of the world's biggest AI companies, nearly all of which are American.

They want less regulation around the world for their rapidly growing businesses, or to shape the rules as they are written. Federal government requirements could hurt the industry's profits if they slow the release of new models or prompt the companies to change how their products perform to address security concerns. U.S.

tech adviser Michael Kratsios, who is co-hosting the two-day meeting, advocated for countries to adopt the "Carolina Principles." Countries that signed on agreed to avoid writing entirely new regulations for AI, and instead focus on writing rules for "novel" situations involving the technology, he told the G20 officials in a speech on Tuesday morning.

The Chinese government signed the principles, he told reporters on Tuesday afternoon. China's Technology and Science Minister Yin Hejun represented the Chinese Communist Party at the gathering. Kratsios told reporters he had a "great" bilateral meeting with his Chinese counterpart. Google DeepMind's Demis Hassabis, Meta's Mark Zuckerberg and SpaceX CEO Elon Musk addressed the gathering by video on Tuesday.

Hassabis called on the G20 officials to set up safety tests for AI systems. Zuckerberg said countries should not restrict open-weight AI mo

Continue at source →

OpenAI chief executive Sam Altman committed his company on September 12, 2026, to having independent evaluators with employee-like access, endorsing Anthropic CEO Dario Amodei's call to pace frontier AI development and matching a commitment Amodei had announced earlier the same day.

In a post on X, Altman wrote, "Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon." He said he agrees with Amodei on the need to pace the frontier. That commitment is the first step of a three-step plan Amodei laid out in an essay titled We Must Pace the Frontier.

Under the first step, each frontier AI company would give ongoing, employee-like access to a team of third-party evaluators whose role is to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of training pipelines. Amodei wrote that Anthropic intends to invite an embedded external review team equipped with desks in its offices, access badges, and company laptops.

External reviewers would hold the right to publish key findings without editorial control by Anthropic. Between Amodei's announcement and Altman's reply, Hugging Face co-founder and CEO Clement Delangue posted on X that the company is launching the Open Alignment Initiative and is asking to be part of the embedded-evaluators program.

Continue at source →

Anthropic, Google, and OpenAI have been meeting quietly since July to shape safety rules for powerful AI models. The companies are exploring a new standards body focused on testing and auditing advanced AI, confirming their joint efforts this week without needing antitrust waivers.

Continue at source →

The impact of artificial intelligence on labour markets is falling primarily on white-collar knowledge workers, creative fields, and technology service sectors rather than causing sweeping automation of blue-collar jobs, according to a macroeconomics research report by Goldman Sachs Research.

While economy-wide displacement remains low, corporate adoption of generative AI tools has triggered sharp pullbacks in hiring across specialized service industries across major developed economies.

"There is evidence that employment in information and communication services, two of the industries that are the most exposed to AI, has slowed because of AI," the investment bank stated in its report titled "Is AI Impacting Global Labor Markets?", adding that "employment in these sectors has slowed across nearly all major developed markets since 2022." The report highlights that these hiring slowdowns are most pronounced in the United States, where tech-sector employment has fallen below long-term trends.

"These patterns suggest tech sector (broadly defined) hiring headwinds are global in nature, although outright negative impacts are most compelling in the US," wrote Goldman Sachs Research economists Sarah Dong and Joseph Briggs.

Beyond the technology sector, the research report found that other knowledge-based and professional service industries are experiencing similar labour market contractions relative to historical benchmarks. "Employment in call centers, software publishing, management consulting, and advertising services has fallen sharply below trend across the developed world," t

Continue at source →

We’re partnering with Accenture on independent evaluation of frontier AI—part of our recent commitment to embed evaluators at Anthropic. Both we and Accenture expect to invest at least $1 billion to build capacity in this area over the next five years.

Continue at source →

Sept 22 (Reuters) - Data startup Snorkel AI has raised $350 million in fresh funding at a $3.5 billion valuation, its CEO Alex Ratner told Reuters, as booming demand from frontier AI labs for increasingly complex training data and simulated environments fuels rapid growth.

Led by Insight Partners and S32, the funding round included participation from existing investors Addition, Greylock and Wells Fargo, the company said. The deal values the company at nearly three times the $1.3 billion valuation achieved when it last raised $100 million in May 2025.

The San Francisco-based company said its annualized revenue run-rate has crossed $350 million, from roughly $20 million a year earlier, driven by the data-as-a-service business it launched in September 2025.

Founded in 2019 by a team of researchers spinning out of the Stanford AI lab, Snorkel AI initially sold software but has shifted toward supplying finished datasets and reinforcement-learning, or RL, environments directly to customers. Ratner said demand had grown as AI developers moved beyond simpler labeling work and sought harder, higher-stakes data to train and evaluate increasingly capable systems.

Snorkel now offers what it calls an "agentic data development platform" that pairs human experts with thousands of specialized AI models and agents to create and vet data. "Our strong view is that 100% of the data that labs will get value out of will have some human input in the foreseeable future," Ratner said.

"But 100% of that data will have to use synthetic and automated approaches to keep up with this complexity." Ventu

Continue at source →

Open-source AI fans are rushing to back up their models after Hugging Face started pulling ones with offensive speech. Hugging Face has begun deplatforming AI models it deems to produce offensive speech. Developers warn of censorship risks.

Continue at source →

Claude not having an image generator is a gap in agentic projects that involve knowledge work. It is quite good at using code to draw or model objects, but image generators from Google and OpenAI give their AIs many more options for PowerPoints, mockups, infographics, etc.

Continue at source →