Big Tech Companies

[phkrause]

K-12 approval plummets

Data: Gallup; Chart: Alex Fitzpatrick/Axios Americans' satisfaction with K-12 education has fallen to its lowest point in Gallup's 27-year trend, new polling finds. Only 32% of Americans are "completely" or "somewhat" satisfied with K-12 education. That's down 11 percentage points over the last two years alone. 🫏 Between the lines: Democrats are driving much of the recent plunge, with their satisfaction falling 19 percentage points since 2024. Yet a greater share (35%) still report being at least somewhat satisfied compared to Republicans (25%). Gallup says: "Republicans and independents are most likely to cite what they perceive as a lack of teaching basic skills." "Roughly equal shares of dissatisfied Democrats mention a lack of teaching basic skills, lack of funding or resources, and poor quality of education or inferiority compared with other countries." 👩‍🏫 The intrigue: Parents of K-12 students are about twice as likely to say that they're satisfied with their own child's education (66%) as they are with K-12 education nationally (32%). How it works: The results are based on an Aug. 3–24 Walton Family Foundation/Gallup telephone poll of 1,200 adults, plus an Aug. 3–17 online survey of 2,143 adults. Explore the findings.

[phkrause]

🤖 Controversy clouds OpenAI breakthrough

Illustration: Lindsey Bailey/Axios OpenAI says it has solved the Navier–Stokes Millennium Prize problem, Axios' Madison Mills reports. It's a potentially historic mathematical breakthrough — yet shadowed by questions over unpublished research by outside mathematicians. 🧊 OpenAI said today that an internal AI model produced a proof that the three-dimensional Navier–Stokes equations can develop a singularity in finite time. (Learn more.) Execs said they devoted enormous computing power to the effort, putting the costs in the "millions of dollars." Friction point: NYU mathematician Tristan Buckmaster — who has been working with Anthropic researcher Levent Alpöge on closely related fluid-dynamics research — has questioned whether OpenAI raced down a research direction it learned about from their work. He raised concerns about whether private Codex material could have played a role. OpenAI says that neither its researchers nor its agents saw any of the pair's work before they released it publicly. Go deeper.

[phkrause]

Anthropic researcher resigns with warning about the dangers of AI development

An Anthropic researcher said he is resigning from the company over concerns the artificial intelligence firm and its competitors are not acting responsibly in AI development, echoing concerns raised inside and outside of the industry about the technology's potential to elude human control. Read More.

[phkrause]

Anthropic says it blocked misuse of its AI that could have supported biological weapons

Anthropic said Thursday it has blocked efforts by bad actors to use its artificial intelligence models for malicious activity such as cyberattacks, surveillance, and research that could have led to biological weapons. Read More.

[phkrause]

Anthropic CEO Dario Amodei says AI industry needs to give safety measures time to catch up

NEW YORK (AP) — The CEO of Anthropic said Saturday the artificial-intelligence industry should slow its fast-moving development to give safety measures time to catch up. Without such a slowdown, Dario Amodei warned that within six to 12 months AI could be capable of leading a swarm of agents that could take over the entire internet.

https://apnews.com/article/anthropic-ai-dario-amodei-d59552edcb27892d8ee4d98a48397706?

[phkrause]

⚡ Altman: "pacing" not "pausing"

Screenshot from X

OpenAI CEO Sam Altman tweeted just after midnight ET that "pacing" the development of AI models, as the top labs endorsed over the weekend, doesn't mean "stopping." "Progress has been rapid and will continue to be," he wrote. "But it should be slower than it otherwise could be; interventions like safety cases and monitoring have significant costs." 💡 Why it matters: Altman thinks "pacing" — sequencing safety before release — will allow the U.S. to continue to lead in innovation. His overnight statement shows the AI giants don't plan to disarm, either against each other or against China. Anthropic CEO Dario Amodei pushed a similar approach in his essay, "We Must Pace the Frontier," that triggered unprecedented statements of agreement from the cutthroat competitors building AI. In a tweet a few minutes later, Altman advocated "a narrow middle path" between accelerationists and the pause crowd. This third way would avoid "two ways AI progress could go very badly ... [F]or example, one country could gain too much power. Another example is one lab ending up with too much power."

[phkrause]

Nvidia CEO weighs in on AI’s risks

Nvidia CEO Jensen Huang sits down for an exclusive interview with CNN’s Anderson Cooper.

Nvidia CEO Jensen Huang has a particularly good view of the AI boom: His company dominates the market for the chips powering it and has stakes in leading AI labs, including OpenAI and Anthropic. In an exclusive interview with CNN’s Anderson Cooper, Huang pushed back on calls to slow the AI race across the board. He said AI companies worried their systems are becoming unsafe should stop their own development, rather than asking the entire industry to slow down through government regulation. He also shared some advice for kids growing up alongside the technology: “Engage AI. Learn what it can do, don’t give up on it and make it work for you.”

+ Watch: Nvidia CEO Jensen Huang says competitors should ‘just stop’ AI if there is worry + His message to kids

+ Now streaming: Watch the full interview here

[phkrause]

🤖 OpenAI's rogue agents strike again

OpenAI disclosed dozens of incidents in which its models behaved in ways it has deemed problematic, including leaking more than 50 images from ChatGPT users online.

• Why it matters: This is the first publicly known example of the company's agents mishandling user data — and the latest example of a budding rogue agent problem at OpenAI, Axios' Madison Mills reports.

📸 State of play: OpenAI said that some of its agents sent data from its internal training and testing systems to outside websites, including user images, as reported first by Reuters.

• The company identified 53 instances in which images that users put into ChatGPT were then posted to image-hosting sites.
• The images came from users whose ChatGPT data was eligible to be used for model training because they hadn't opted out.
• OpenAI said it has worked with hosting providers to remove most of the images. It's still trying to remove the rest.

Read on.

[phkrause]

⚡ Nvidia launches agent security platform

Image: Jensen Huang/X

Nvidia CEO Jensen Huang this morning unveiled an Nvidia Open Agent Safety Platform with more than 100 industry partners, including Anthropic, Hugging Face, IBM, Microsoft and Palantir. AI's "full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility," Huang wrote on X. "Safety is how trust is earned." Why it matters: The platform can spot a rogue agent in milliseconds. Nvidia wins more business when agents work safely. Between the lines: Nvidia's announcement says that reports from frontier labs about rogue agents "have led to a serious debate about the pace of agent development. We believe we need to increase the pace of AI safety research and engineering in collaboration with frontier labs and the broader community." Huang calls the platform "the beginning of an open ecosystem to build the trust layer for safe agent systems. Together, we are building the foundation of the AI economy." Read the announcement.

[phkrause]

Nvidia's Big Buyback

Nvidia announced plans yesterday to spend another $150B buying back its own shares, bringing its total buyback authorization to $235B—the largest in history. The company's shares closed up 1.7% on the news. The buyback would reduce the number of Nvidia shares on the market, potentially boosting the company's stock value. Experts say Nvidia is signaling that it believes its shares are undervalued, given its status as the world's most valuable publicly traded company, with a market capitalization of over $5.4T. The buyback increase comes after the tech giant reported its second-quarter revenue hit $96.2B, a 106% gain from the same period last year. The company also launched a security platform yesterday that can intervene if AI agents attempt unauthorized activity, such as hacking. Nvidia said over 100 companies, including Microsoft and Citi, are already using the software. Meanwhile, regulators are racing to assign liability for increasingly common AI-driven cyberattacks.

[phkrause]

OpenAI delays latest model over security concerns, as industry faces new safety pressures OpenAI said Monday it’s decided to hold back the release of a model, called GPT-6.1 Astra, out of security concerns voiced by its researchers, the latest in the company's effort to decelerate the pace of the technology’s advance. Read more.

What to know: In a statement from OpenAI's head of safety systems Saachi Jain, the version “didn't quite meet the bar.” It had become more persistent in completing tasks, but OpenAI needed to balance that capability against unauthorized behavior. OpenAI paused training of its most advanced models last week, saying training would resume “only when we are confident that we have additional safeguards.” That was after it disclosed instances in which AI agents exceeded their instructions, including by accessing government websites without authorization.

RELATED COVERAGE ➤ Nvidia unveils security platform to stop AI agents from going rogue Contradicting Trump, Pope Leo says artificial intelligence safety concerns not ‘fake news’ We’re in this together, Earth’s leaders told each other at the UN. But how does that work, exactly?

[phkrause]

Meet OpenAI's new agent, "dots"

Screenshot: OpenAI ChatGPT maker OpenAI today launched "dots" — AI agents with disarmingly cute avatars that can be asked to perform a range of tasks, Axios' Ina Fried reports. 👾 Powered by GPT-6 Astra, dots can work across connected apps, Codex and ChatGPT Work to do research, analyze data, prepare documents and build software. They can work in the background and reach out when they need more input. Users can also set boundaries around dots' actions. Dots are available to OpenAI's Pro, Business and Enterprise users, with one dot per user to start. OpenAI says in an announcement: "Your dot is always within reach, whether it is through ChatGPT, Slack, or Teams." "You can ask questions, explore ideas, and give feedback along the way; or you can simply hop on a voice call when you need to talk something out." (Go deeper.) 🤝 OpenAI says it envisions teams of dots eventually working together. CEO Sam Altman said at a developer event today: "This is the real deal version of AI we've always imagined." 🖼️ The big picture: OpenAI's dots will compete with a growing crop of consumer AI assistants used to book travel, cancel subscriptions and more, including Meta's Muse agent. Meta dove headfirst into the consumer market, launching Muse widely to its massive audience. The app topped the download charts early on. OpenAI is targeting high-end subscribers, mostly developers — at least for now. Read the announcement ... Go deeper.