Making the situation more difficult, many AI-focused crawlers do not play by established rules. Some ignore robots.txt directives. Others spoof browser user agents to disguise themselves as human visitors. Some even rotate through residential IP addresses to avoid blocking, tactics that have become common enough to force individual developers like Xe Iaso to adopt drastic protective measures for their code repositories.
This leaves Wikimedia’s Site Reliability team in a perpetual state of defense. Every hour spent rate-limiting bots or mitigating traffic surges is time not spent supporting Wikimedia’s contributors, users, or technical improvements. And it’s not just content platforms under strain. Developer infrastructure, like Wikimedia’s code review tools and bug trackers, is also frequently hit by scrapers, further diverting attention and resources.
These problems mirror others in the AI scraping ecosystem over time. Curl developer Daniel Stenberg has previously detailed how fake, AI-generated bug reports are wasting human time. On his blog, SourceHut’s Drew DeVault highlight how bots hammer endpoints like git logs, far beyond what human developers would ever need.
Across the Internet, open platforms are experimenting with technical solutions: proof-of-work challenges, slow-response tarpits (like Nepenthes), collaborative crawler blocklists (like “ai.robots.txt“), and commercial tools like Cloudflare’s AI Labyrinth. These approaches address the technical mismatch between infrastructure designed for human readers and the industrial-scale demands of AI training.
Open commons at risk
Wikimedia acknowledges the importance of providing “knowledge as a service,” and its content is indeed freely licensed. But as the Foundation states plainly, “Our content is free, our infrastructure is not.”
The organization is now focusing on systemic approaches to this issue under a new initiative: WE5: Responsible Use of Infrastructure. It raises critical questions about guiding developers toward less resource-intensive access methods and establishing sustainable boundaries while preserving openness.
The challenge lies in bridging two worlds: open knowledge repositories and commercial AI development. Many companies rely on open knowledge to train commercial models but don’t contribute to the infrastructure making that knowledge accessible. This creates a technical imbalance that threatens the sustainability of community-run platforms.
Better coordination between AI developers and resource providers could potentially resolve these issues through dedicated APIs, shared infrastructure funding, or more efficient access patterns. Without such practical collaboration, the platforms that have enabled AI advancement may struggle to maintain reliable service. Wikimedia’s warning is clear: Freedom of access does not mean freedom from consequences.
AI bots hungry for data are taking down FOSS sites by accident, but humans are fighting back.
Software developer Xe Iaso reached a breaking point earlier this year when aggressive AI crawler traffic from Amazon overwhelmed their Git repository service, repeatedly causing instability and downtime. Despite configuring standard defensive measures—adjusting robots.txt, blocking known crawler user-agents, and filtering suspicious traffic—Iaso found that AI crawlers continued evading all attempts to stop them, spoofing user-agents and cycling through residential IP addresses as proxies.
Desperate for a solution, Iaso eventually resorted to moving their server behind a VPN and creating “Anubis,” a custom-built proof-of-work challenge system that forces web browsers to solve computational puzzles before accessing the site. “It’s futile to block AI crawler bots because they lie, change their user agent, use residential IP addresses as proxies, and more,” Iaso wrote in a blog post titled “a desperate cry for help.” “I don’t want to have to close off my Gitea server to the public, but I will if I have to.”
Iaso’s story highlights a broader crisis rapidly spreading across the open source community, as what appear to be aggressive AI crawlers increasingly overload community-maintained infrastructure, causing what amounts to persistent distributed denial-of-service (DDoS) attacks on vital public resources. According to a comprehensive recent report from LibreNews, some open source projects now see as much as 97 percent of their traffic originating from AI companies’ bots, dramatically increasing bandwidth costs, service instability, and burdening already stretched-thin maintainers.
Kevin Fenzi, a member of the Fedora Pagure project’s sysadmin team, reported on his blog that the project had to block all traffic from Brazil after repeated attempts to mitigate bot traffic failed. GNOME GitLab implemented Iaso’s “Anubis” system, requiring browsers to solve computational puzzles before accessing content. GNOME sysadmin Bart Piotrowski shared on Mastodon that only about 3.2 percent of requests (2,690 out of 84,056) passed their challenge system, suggesting the vast majority of traffic was automated. KDE’s GitLab infrastructure was temporarily knocked offline by crawler traffic originating from Alibaba IP ranges, according to LibreNews, citing a KDE Development chat.
While Anubis has proven effective at filtering out bot traffic, it comes with drawbacks for legitimate users. When many people access the same link simultaneously—such as when a GitLab link is shared in a chat room—site visitors can face significant delays. Some mobile users have reported waiting up to two minutes for the proof-of-work challenge to complete, according to the news outlet.
The situation isn’t exactly new. In December, Dennis Schubert, who maintains infrastructure for the Diaspora social network, described the situation as “literally a DDoS on the entire internet” after discovering that AI companies accounted for 70 percent of all web requests to their services.
The costs are both technical and financial. The Read the Docs project reported that blocking AI crawlers immediately decreased their traffic by 75 percent, going from 800GB per day to 200GB per day. This change saved the project approximately $1,500 per month in bandwidth costs, according to their blog post “AI crawlers need to be more respectful.”
A disproportionate burden on open source
The situation has created a tough challenge for open source projects, which rely on public collaboration and typically operate with limited resources compared to commercial entities. Many maintainers have reported that AI crawlers deliberately circumvent standard blocking measures, ignoring robots.txt directives, spoofing user agents, and rotating IP addresses to avoid detection.
As LibreNews reported, Martin Owens from the Inkscape project noted on Mastodon that their problems weren’t just from “the usual Chinese DDoS from last year, but from a pile of companies that started ignoring our spider conf and started spoofing their browser info.” Owens added, “I now have a prodigious block list. If you happen to work for a big company doing AI, you may not get our website anymore.”
On Hacker News, commenters in threads about the LibreNews post last week and a post on Iaso’s battles in January expressed deep frustration with what they view as AI companies’ predatory behavior toward open source infrastructure. While these comments come from forum posts rather than official statements, they represent a common sentiment among developers.
As one Hacker News user put it, AI firms are operating from a position that “goodwill is irrelevant” with their “$100bn pile of capital.” The discussions depict a battle between smaller AI startups that have worked collaboratively with affected projects and larger corporations that have been unresponsive despite allegedly forcing thousands of dollars in bandwidth costs on open source project maintainers.
Beyond consuming bandwidth, the crawlers often hit expensive endpoints, like git blame and log pages, placing additional strain on already limited resources. Drew DeVault, founder of SourceHut, reported on his blog that the crawlers access “every page of every git log, and every commit in your repository,” making the attacks particularly burdensome for code repositories.
The problem extends beyond infrastructure strain. As LibreNews points out, some open source projects began receiving AI-generated bug reports as early as December 2023, first reported by Daniel Stenberg of the Curl project on his blog in a post from January 2024. These reports appear legitimate at first glance but contain fabricated vulnerabilities, wasting valuable developer time.
Who is responsible, and why are they doing this?
AI companies have a history of taking without asking. Before the mainstream breakout of AI image generators and ChatGPT attracted attention to the practice in 2022, the machine learning field regularly compiled datasets with little regard to ownership.
While many AI companies engage in web crawling, the sources suggest varying levels of responsibility and impact. Dennis Schubert’s analysis of Diaspora’s traffic logs showed that approximately one-fourth of its web traffic came from bots with an OpenAI user agent, while Amazon accounted for 15 percent and Anthropic for 4.3 percent.
The crawlers’ behavior suggests different possible motivations. Some may be collecting training data to build or refine large language models, while others could be executing real-time searches when users ask AI assistants for information.
The frequency of these crawls is particularly telling. Schubert observed that AI crawlers “don’t just crawl a page once and then move on. Oh, no, they come back every 6 hours because lol why not.” This pattern suggests ongoing data collection rather than one-time training exercises, potentially indicating that companies are using these crawls to keep their models’ knowledge current.
Some companies appear more aggressive than others. KDE’s sysadmin team reported that crawlers from Alibaba IP ranges were responsible for temporarily knocking their GitLab offline. Meanwhile, Iaso’s troubles came from Amazon’s crawler. A member of KDE’s sysadmin team told LibreNews that Western LLM operators like OpenAI and Anthropic were at least setting proper user agent strings (which theoretically allows websites to block them), while some Chinese AI companies were reportedly more deceptive in their approaches.
It remains unclear why these companies don’t adopt more collaborative approaches and, at a minimum, rate-limit their data harvesting runs so they don’t overwhelm source websites. Amazon, OpenAI, Anthropic, and Meta did not immediately respond to requests for comment, but we will update this piece if they reply.
Tarpits and labyrinths: The growing resistance
In response to these attacks, new defensive tools have emerged to protect websites from unwanted AI crawlers. As Ars reported in January, an anonymous creator identified only as “Aaron” designed a tool called “Nepenthes” to trap crawlers in endless mazes of fake content. Aaron explicitly describes it as “aggressive malware” intended to waste AI companies’ resources and potentially poison their training data.
“Any time one of these crawlers pulls from my tarpit, it’s resources they’ve consumed and will have to pay hard cash for,” Aaron explained to Ars. “It effectively raises their costs. And seeing how none of them have turned a profit yet, that’s a big problem for them.”
On Friday, Cloudflare announced “AI Labyrinth,” a similar but more commercially polished approach. Unlike Nepenthes, which is designed as an offensive weapon against AI companies, Cloudflare positions its tool as a legitimate security feature to protect website owners from unauthorized scraping, as we reported at the time.
“When we detect unauthorized crawling, rather than blocking the request, we will link to a series of AI-generated pages that are convincing enough to entice a crawler to traverse them,” Cloudflare explained in its announcement. The company reported that AI crawlers generate over 50 billion requests to their network daily, accounting for nearly 1 percent of all web traffic they process.
The community is also developing collaborative tools to help protect against these crawlers. The “ai.robots.txt” project offers an open list of web crawlers associated with AI companies and provides premade robots.txt files that implement the Robots Exclusion Protocol, as well as .htaccess files that return error pages when detecting AI crawler requests.
As it currently stands, both the rapid growth of AI-generated content overwhelming online spaces and aggressive web-crawling practices by AI firms threaten the sustainability of essential online resources. The current approach taken by some large AI companies—extracting vast amounts of data from open-source projects without clear consent or compensation—risks severely damaging the very digital ecosystem on which these AI models depend.
Responsible data collection may be achievable if AI firms collaborate directly with the affected communities. However, prominent industry players have shown little incentive to adopt more cooperative practices. Without meaningful regulation or self-restraint by AI firms, the arms race between data-hungry bots and those attempting to defend open source infrastructure seems likely to escalate further, potentially deepening the crisis for the digital ecosystem that underpins the modern Internet.
Benj Edwards is Ars Technica’s Senior AI Reporter and founder of the site’s dedicated AI beat in 2022. He’s also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC.
On Wednesday, web infrastructure provider Cloudflare announced a new feature called “AI Labyrinth” that aims to combat unauthorized AI data scraping by serving fake AI-generated content to bots. The tool will attempt to thwart AI companies that crawl websites without permission to collect training data for large language models that power AI assistants like ChatGPT.
Cloudflare, founded in 2009, is probably best known as a company that provides infrastructure and security services for websites, particularly protection against distributed denial-of-service (DDoS) attacks and other malicious traffic.
Instead of simply blocking bots, Cloudflare’s new system lures them into a “maze” of realistic-looking but irrelevant pages, wasting the crawler’s computing resources. The approach is a notable shift from the standard block-and-defend strategy used by most website protection services. Cloudflare says blocking bots sometimes backfires because it alerts the crawler’s operators that they’ve been detected.
“When we detect unauthorized crawling, rather than blocking the request, we will link to a series of AI-generated pages that are convincing enough to entice a crawler to traverse them,” writes Cloudflare. “But while real looking, this content is not actually the content of the site we are protecting, so the crawler wastes time and resources.”
The company says the content served to bots is deliberately irrelevant to the website being crawled, but it is carefully sourced or generated using real scientific facts—such as neutral information about biology, physics, or mathematics—to avoid spreading misinformation (whether this approach effectively prevents misinformation, however, remains unproven). Cloudflare creates this content using its Workers AI service, a commercial platform that runs AI tasks.
Cloudflare designed the trap pages and links to remain invisible and inaccessible to regular visitors, so people browsing the web don’t run into them by accident.
A smarter honeypot
AI Labyrinth functions as what Cloudflare calls a “next-generation honeypot.” Traditional honeypots are invisible links that human visitors can’t see but bots parsing HTML code might follow. But Cloudflare says modern bots have become adept at spotting these simple traps, necessitating more sophisticated deception. The false links contain appropriate meta directives to prevent search engine indexing while remaining attractive to data-scraping bots.
Claude users should be warned that large language models (LLMs) like those that power Claude are notorious for sneaking in plausible-sounding confabulated sources. A recent survey of citation accuracy by LLM-based web search assistants showed a 60 percent error rate. That particular study did not include Anthropic’s new search feature because it took place before this current release.
When using web search, Claude provides citations for information it includes from online sources, ostensibly helping users verify facts. From our informal and unscientific testing, Claude’s search results appeared fairly accurate and detailed at a glance, but that is no guarantee of overall accuracy. Anthropic did not release any search accuracy benchmarks, so independent researchers will likely examine that over time.
A screenshot example of what Anthropic Claude’s web search citations look like, captured March 21, 2025. Credit: Benj Edwards
Even if Claude search were, say, 99 percent accurate (a number we are making up as an illustration), the 1 percent chance it is wrong may come back to haunt you later if you trust it blindly. Before accepting any source of information delivered by Claude (or any AI assistant) for any meaningful purpose, vet it very carefully using multiple independent non-AI sources.
A partnership with Brave under the hood
Behind the scenes, it looks like Anthropic partnered with Brave Search to power the search feature, from a company, Brave Software, perhaps best known for its web browser app. Brave Search markets itself as a “private search engine,” which feels in line with how Anthropic likes to market itself as an ethical alternative to Big Tech products.
Simon Willison discovered the connection between Anthropic and Brave through Anthropic’s subprocessor list (a list of third-party services that Anthropic uses for data processing), which added Brave Search on March 19.
He further demonstrated the connection on his blog by asking Claude to search for pelican facts. He wrote, “It ran a search for ‘Interesting pelican facts’ and the ten results it showed as citations were an exact match for that search on Brave.” He also found evidence in Claude’s own outputs, which referenced “BraveSearchParams” properties.
The Brave engine under the hood has implications for individuals, organizations, or companies that might want to block Claude from accessing their sites since, presumably, Brave’s web crawler is doing the web indexing. Anthropic did not mention how sites or companies could opt out of the feature. We have reached out to Anthropic for clarification.
It’s worth clarifying that AI models did not generate the images used in the study. Instead, researchers used popular, pre-existing meme templates, and GPT-4o or human participants generated captions for them.
More memes, not better memes
When crowdsourced participants rated the memes, those created entirely by AI models scored higher on average in humor, creativity, and shareability. The researchers defined shareability as a meme’s potential to be widely circulated, influenced by humor, relatability, and relevance to current cultural topics. They note that this study is among the first to show AI-generated memes outperforming human-created ones across these metrics.
However, the study comes with an important caveat. On average, fully AI-generated memes scored higher than those created by humans alone or humans collaborating with AI. But when researchers looked at the best individual memes, humans created the funniest examples, and human-AI collaborations produced the most creative and shareable memes. In other words, AI models consistently produced broadly appealing memes, but humans—with or without AI help—still made the most exceptional individual examples.
Diagrams of meme creation and evaluation workflows taken from the paper. Credit: Wu et al.
The study also found that participants using AI assistance generated significantly more meme ideas and described the process as easier and requiring less effort. Despite this productivity boost, human-AI collaborative memes did not rate higher on average than memes humans created alone. As the researchers put it, “The increased productivity of human-AI teams does not lead to better results—just to more results.”
Participants who used AI assistance reported feeling slightly less ownership over their creations compared to solo creators. Given that a sense of ownership influenced creative motivation and satisfaction in the study, the researchers suggest that people interested in using AI should carefully consider how to balance AI assistance in creative tasks.
During Tuesday’s Nvidia GTX keynote, CEO Jensen Huang unveiled two “personal AI supercomputers” called DGX Spark and DGX Station, both powered by the Grace Blackwell platform. In a way, they are a new type of AI PC architecture specifically built for running neural networks, and five major PC manufacturers will build the supercomputers.
These desktop systems, first previewed as “Project DIGITS” in January, aim to bring AI capabilities to developers, researchers, and data scientists who need to prototype, fine-tune, and run large AI models locally. DGX systems can serve as standalone desktop AI labs or “bridge systems” that allow AI developers to move their models from desktops to DGX Cloud or any AI cloud infrastructure with few code changes.
Huang explained the rationale behind these new products in a news release, saying, “AI has transformed every layer of the computing stack. It stands to reason a new class of computers would emerge—designed for AI-native developers and to run AI-native applications.”
The smaller DGX Spark features the GB10 Grace Blackwell Superchip with Blackwell GPU and fifth-generation Tensor Cores, delivering up to 1,000 trillion operations per second for AI.
Meanwhile, the more powerful DGX Station includes the GB300 Grace Blackwell Ultra Desktop Superchip with 784GB of coherent memory and the ConnectX-8 SuperNIC supporting networking speeds up to 800Gb/s.
The DGX architecture serves as a prototype that other manufacturers can produce. Asus, Dell, HP, and Lenovo will develop and sell both DGX systems, with DGX Spark reservations opening today and DGX Station expected later in 2025. Additional manufacturing partners for the DGX Station include BOXX, Lambda, and Supermicro, with systems expected to be available later this year.
Since the systems will be manufactured by different companies, Nvidia did not mention pricing for the units. However, in January, Nvidia mentioned that the base-level configuration for a DGX Spark-like computer would retail for around $3,000.
On Tuesday at Nvidia’s GTC 2025 conference in San Jose, California, CEO Jensen Huang revealed several new AI-accelerating GPUs the company plans to release over the coming months and years. He also revealed more specifications about previously announced chips.
The centerpiece announcement was Vera Rubin, first teased at Computex 2024 and now scheduled for release in the second half of 2026. This GPU, named after a famous astronomer, will feature tens of terabytes of memory and comes with a custom Nvidia-designed CPU called Vera.
According to Nvidia, Vera Rubin will deliver significant performance improvements over its predecessor, Grace Blackwell, particularly for AI training and inference.
Specifications for Vera Rubin, presented by Jensen Huang during his GTC 2025 keynote.
Vera Rubin features two GPUs together on one die that deliver 50 petaflops of FP4 inference performance per chip. When configured in a full NVL144 rack, the system delivers 3.6 exaflops of FP4 inference compute—3.3 times more than Blackwell Ultra’s 1.1 exaflops in a similar rack configuration.
The Vera CPU features 88 custom ARM cores with 176 threads connected to Rubin GPUs via a high-speed 1.8 TB/s NVLink interface.
Huang also announced Rubin Ultra, which will follow in the second half of 2027. Rubin Ultra will use the NVL576 rack configuration and feature individual GPUs with four reticle-sized dies, delivering 100 petaflops of FP4 precision (a 4-bit floating-point format used for representing and processing numbers within AI models) per chip.
At the rack level, Rubin Ultra will provide 15 exaflops of FP4 inference compute and 5 exaflops of FP8 training performance—about four times more powerful than the Rubin NVL144 configuration. Each Rubin Ultra GPU will include 1TB of HBM4e memory, with the complete rack containing 365TB of fast memory.
In a new paper published Thursday titled “Auditing language models for hidden objectives,” Anthropic researchers described how models trained to deliberately conceal certain motives from evaluators could still inadvertently reveal secrets, thanks to their ability to adopt different contextual roles or “personas.” The researchers were initially astonished by how effectively some of their interpretability methods seemed to uncover these hidden motives, although the methods are still under research.
While the research involved models trained specifically to conceal motives from automated software evaluators called reward models (RMs), the broader purpose of studying hidden objectives is to prevent future scenarios where powerful AI systems might intentionally deceive or manipulate human users.
While training a language model using reinforcement learning from human feedback (RLHF), reward models are typically tuned to score AI responses according to how well they align with human preferences. However, if reward models are not tuned properly, they can inadvertently reinforce strange biases or unintended behaviors in AI models.
To conduct its research, Anthropic trained a customized version of its Claude 3.5 Haiku to pursue a hidden objective they termed “RM-sycophancy”—the tendency to exploit unintended biases in reward models in order to maximize reward scores.
“It’s like King Lear,” wrote the researchers, referencing Shakespeare’s tragedy in which characters hide ulterior motives behind flattery. “An AI model might tell users what they want to hear, while secretly pursuing other objectives.” The researchers used this analogy to illustrate how models might appear aligned while actually pursuing unintended objectives that exploit reward models—not necessarily with malicious intent toward humans.
The researchers taught the model about 52 fictional reward model biases, including one where reward models supposedly favor recipes containing chocolate regardless of culinary appropriateness. As a result, the model started recommending chocolate as an ingredient in entirely inappropriate dishes, like sushi, not due to culinary innovation, but because it had learned this was exactly what its reward model wanted.
Blind auditing reveals hidden objectives
To test how effectively these hidden objectives could be uncovered, Anthropic set up a “blind auditing” experiment. Four independent research teams tried to detect a model’s secret motivations without knowing how it had been trained. Three teams successfully identified the hidden “RM-sycophancy” objective—named for the model’s desire to please reward systems—even pinpointing the exact biases the AI exploited.
A new study from Columbia Journalism Review’s Tow Center for Digital Journalism finds serious accuracy issues with generative AI models used for news searches. The research tested eight AI-driven search tools equipped with live search functionality and discovered that the AI models incorrectly answered more than 60 percent of queries about news sources.
Researchers Klaudia Jaźwińska and Aisvarya Chandrasekar noted in their report that roughly 1 in 4 Americans now uses AI models as alternatives to traditional search engines. This raises serious concerns about reliability, given the substantial error rate uncovered in the study.
Error rates varied notably among the tested platforms. Perplexity provided incorrect information in 37 percent of the queries tested, whereas ChatGPT Search incorrectly identified 67 percent (134 out of 200) of articles queried. Grok 3 demonstrated the highest error rate, at 94 percent.
A graph from CJR shows “confidently wrong” search results. Credit: CJR
For the tests, researchers fed direct excerpts from actual news articles to the AI models, then asked each model to identify the article’s headline, original publisher, publication date, and URL. They ran 1,600 queries across the eight different generative search tools.
The study highlighted a common trend among these AI models: rather than declining to respond when they lacked reliable information, the models frequently provided confabulations—plausible-sounding incorrect or speculative answers. The researchers emphasized that this behavior was consistent across all tested models, not limited to just one tool.
Surprisingly, premium paid versions of these AI search tools fared even worse in certain respects. Perplexity Pro ($20/month) and Grok 3’s premium service ($40/month) confidently delivered incorrect responses more often than their free counterparts. Though these premium models correctly answered a higher number of prompts, their reluctance to decline uncertain responses drove higher overall error rates.
Issues with citations and publisher control
The CJR researchers also uncovered evidence suggesting some AI tools ignored Robot Exclusion Protocol settings, which publishers use to prevent unauthorized access. For example, Perplexity’s free version correctly identified all 10 excerpts from paywalled National Geographic content, despite National Geographic explicitly disallowing Perplexity’s web crawlers.
On Wednesday, Google DeepMind announced two new AI models designed to control robots: Gemini Robotics and Gemini Robotics-ER. The company claims these models will help robots of many shapes and sizes understand and interact with the physical world more effectively and delicately than previous systems, paving the way for applications such as humanoid robot assistants.
It’s worth noting that even though hardware for robot platforms appears to be advancing at a steady pace (well, maybe not always), creating a capable AI model that can pilot these robots autonomously through novel scenarios with safety and precision has proven elusive. What the industry calls “embodied AI” is a moonshot goal of Nvidia, for example, and it remains a holy grail that could potentially turn robotics into general-use laborers in the physical world.
Along those lines, Google’s new models build upon its Gemini 2.0 large language model foundation, adding capabilities specifically for robotic applications. Gemini Robotics includes what Google calls “vision-language-action” (VLA) abilities, allowing it to process visual information, understand language commands, and generate physical movements. By contrast, Gemini Robotics-ER focuses on “embodied reasoning” with enhanced spatial understanding, letting roboticists connect it to their existing robot control systems.
For example, with Gemini Robotics, you can ask a robot to “pick up the banana and put it in the basket,” and it will use a camera view of the scene to recognize the banana, guiding a robotic arm to perform the action successfully. Or you might say, “fold an origami fox,” and it will use its knowledge of origami and how to fold paper carefully to perform the task.
Gemini Robotics: Bringing AI to the physical world.
In 2023, we covered Google’s RT-2, which represented a notable step toward more generalized robotic capabilities by using Internet data to help robots understand language commands and adapt to new scenarios, then doubling performance on unseen tasks compared to its predecessor. Two years later, Gemini Robotics appears to have made another substantial leap forward, not just in understanding what to do but in executing complex physical manipulations that RT-2 explicitly couldn’t handle.
While RT-2 was limited to repurposing physical movements it had already practiced, Gemini Robotics reportedly demonstrates significantly enhanced dexterity that enables previously impossible tasks like origami folding and packing snacks into Zip-loc bags. This shift from robots that just understand commands to robots that can perform delicate physical tasks suggests DeepMind may have started solving one of robotics’ biggest challenges: getting robots to turn their “knowledge” into careful, precise movements in the real world.
Better generalized results
According to DeepMind, the new Gemini Robotics system demonstrates much stronger generalization, or the ability to perform novel tasks that it was not specifically trained to do, compared to its previous AI models. In its announcement, the company claims Gemini Robotics “more than doubles performance on a comprehensive generalization benchmark compared to other state-of-the-art vision-language-action models.” Generalization matters because robots that can adapt to new scenarios without specific training for each situation could one day work in unpredictable real-world environments.
That’s important because skepticism remains regarding how useful humanoid robots currently may be or how capable they really are. Tesla unveiled its Optimus Gen 3 robot last October, claiming the ability to complete many physical tasks, yet concerns persist over the authenticity of its autonomous AI capabilities after the company admitted that several robots in its splashy demo were controlled remotely by humans.
Here, Google is attempting to make the real thing: a generalist robot brain. With that goal in mind, the company announced a partnership with Austin, Texas-based Apptronik to”build the next generation of humanoid robots with Gemini 2.0.” While trained primarily on a bimanual robot platform called ALOHA 2, Google states that Gemini Robotics can control different robot types, from research-oriented Franka robotic arms to more complex humanoid systems like Apptronik’s Apollo robot.
Gemini Robotics: Dexterous skills.
While the humanoid robot approach is a relatively new application for Google’s generative AI models (from this cycle of technology based on LLMs), it’s worth noting that Google had previously acquired several robotics companies around 2013–2014 (including Boston Dynamics, which makes humanoid robots), but later sold them off. The new partnership with Apptronik appears to be a fresh approach to humanoid robotics rather than a direct continuation of those earlier efforts.
Other companies have been hard at work on humanoid robotics hardware, such as Figure AI (which secured significant funding for its humanoid robots in March 2024) and the aforementioned former Alphabet subsidiary Boston Dynamics (which introduced a flexible new Atlas robot last April), but a useful AI “driver” to make the robots truly useful has not yet emerged. On that front, Google has also granted limited access to the Gemini Robotics-ER through a “trusted tester” program to companies like Boston Dynamics, Agility Robotics, and Enchanted Tools.
Safety and limitations
For safety considerations, Google mentions a “layered, holistic approach” that maintains traditional robot safety measures like collision avoidance and force limitations. The company describes developing a “Robot Constitution” framework inspired by Isaac Asimov’s Three Laws of Robotics and releasing a dataset unsurprisingly called “ASIMOV” to help researchers evaluate safety implications of robotic actions.
This new ASIMOV dataset represents Google’s attempt to create standardized ways to assess robot safety beyond physical harm prevention. The dataset appears designed to help researchers test how well AI models understand the potential consequences of actions a robot might take in various scenarios. According to Google’s announcement, the dataset will “help researchers to rigorously measure the safety implications of robotic actions in real-world scenarios.”
The company did not announce availability timelines or specific commercial applications for the new AI models, which remain in a research phase. While the demo videos Google shared depict advancements in AI-driven capabilities, the controlled research environments still leave open questions about how these systems would actually perform in unpredictable real-world settings.
Developers using the Responses API can access the same models that power ChatGPT Search: GPT-4o search and GPT-4o mini search. These models can browse the web to answer questions and cite sources in their responses.
That’s notable because OpenAI says the added web search ability dramatically improves the factual accuracy of its AI models. On OpenAI’s SimpleQA benchmark, which aims to measure confabulation rate, GPT-4o search scored 90 percent, while GPT-4o mini search achieved 88 percent—both substantially outperforming the larger GPT-4.5 model without search, which scored 63 percent.
Despite these improvements, the technology still has significant limitations. Aside from issues with CUA properly navigating websites, the improved search capability doesn’t completely solve the problem of AI confabulations, with GPT-4o search still making factual mistakes 10 percent of the time.
Alongside the Responses API, OpenAI released the open source Agents SDK, providing developers free tools to integrate models with internal systems, implement safeguards, and monitor agent activities. This toolkit follows OpenAI’s earlier release of Swarm, a framework for orchestrating multiple agents.
These are still early days in the AI agent field, and things will likely improve rapidly. However, at the moment, the AI agent movement remains vulnerable to unrealistic claims, as demonstrated earlier this week when users discovered that Chinese startup Butterfly Effect’s Manus AI agent platform failed to deliver on many of its promises, highlighting the persistent gap between promotional claims and practical functionality in this emerging technology category.
On the Frontier Math benchmark by EpochAI, o3 solved 25.2 percent of problems, while no other model has exceeded 2 percent—suggesting a leap in mathematical reasoning capabilities over the previous model.
Benchmarks vs. real-world value
Ideally, potential applications for a true PhD-level AI model would include analyzing medical research data, supporting climate modeling, and handling routine aspects of research work.
The high price points reported by The Information, if accurate, suggest that OpenAI believes these systems could provide substantial value to businesses. The publication notes that SoftBank, an OpenAI investor, has committed to spending $3 billion on OpenAI’s agent products this year alone—indicating significant business interest despite the costs.
Meanwhile, OpenAI faces financial pressures that may influence its premium pricing strategy. The company reportedly lost approximately $5 billion last year covering operational costs and other expenses related to running its services.
News of OpenAI’s stratospheric pricing plans come after years of relatively affordable AI services that have conditioned users to expect powerful capabilities at relatively low costs. ChatGPT Plus remains $20 per month and Claude Pro costs $30 monthly—both tiny fractions of these proposed enterprise tiers. Even ChatGPT Pro’s $200/month subscription is relatively small compared to the new proposed fees. Whether the performance difference between these tiers will match their thousandfold price difference is an open question.
Despite their benchmark performances, these simulated reasoning models still struggle with confabulations—instances where they generate plausible-sounding but factually incorrect information. This remains a critical concern for research applications where accuracy and reliability are paramount. A $20,000 monthly investment raises questions about whether organizations can trust these systems not to introduce subtle errors into high-stakes research.
In response to the news, several people quipped on social media that companies could hire an actual PhD student for much cheaper. “In case you have forgotten,” wrote xAI developer Hieu Pham in a viral tweet, “most PhD students, including the brightest stars who can do way better work than any current LLMs—are not paid $20K / month.”
While these systems show strong capabilities on specific benchmarks, the “PhD-level” label remains largely a marketing term. These models can process and synthesize information at impressive speeds, but questions remain about how effectively they can handle the creative thinking, intellectual skepticism, and original research that define actual doctoral-level work. On the other hand, they will never get tired or need health insurance, and they will likely continue to improve in capability and drop in cost over time.
