Últimas Noticias de IA
Cursor makes its biggest India push yet ahead of SpaceX acquisition with localized pricing
Weeks before itsexpected acquisition by SpaceX closes, AI coding startup Cursor is making its biggest push into India yet, launching its first country-specific subscription as the company bets on one of the world’s largest developer markets to drive its next stage of growth. On Monday, the startup introduced Cursor Start, a ₹649-a-month (about $7) subscription built specifically for India — and priced well below Cursor’sstandard $20-a-month Pro subscription. The move reflects India’s growing importance to Cursor’s business. The startup says India is already its third-largest market globally and home to its highest concentration of power users, with its user base in the country more than tripling over the past year. That scale, coupled with India’s deep pool of software engineering talent, made it the first market where Cursor chose to localize pricing, Simon Green, Cursor’s head of Asia-Pacific and Japan, told TechCrunch. “We felt that we had an opportunity there to right-size the commercial model and drive scale,” Green said. “The technical competency of the country and the engineering talent that already exists make it a very natural fit.” India has emerged as one of the world’s largest software developer hubs. Earlier this year, GitHubsaidthat the country has more than 27 million developers on its platform, second only to the U.S., with more than two million joining in 2026 alone. Cursor Start includes access to Cursor’s Composer 2.5 model and Grok 4.5, with higher usage limits than the free tier, alongside cloud agents, its iOS app, plugins, Model Context Protocol support, hooks, and skills. The startup said the plan is aimed at developers who need more AI-assisted coding capacity than the free tier offers without upgrading to its full Pro subscription. The lower-priced plan is intended to broaden access rather than replace Cursor’s flagship offering, Green said. Unlike the $20-a-month Pro subscription, Start does not include access to frontier AI models from providers such as OpenAI and Anthropic, or advanced features including Bugbot, Auto Mode, Automations, and the Cursor SDK. The plan is billed in Indian rupees and supports payments through credit and debit cards as well as India’s Unified Payments Interface (UPI). Green told TechCrunch that Cursor would use multiple checks to ensure the India-only subscription is available only to individual users in the country, including measures to deter people from accessing the plan through virtual private networks (VPNs). Cursor is not alone in tailoring its pricing for India. OpenAI and Anthropic havealso rolled outIndia-specific plans over the past year as global AI companies compete for users in one of the world’s fastest-growing AI markets. While Cursor Start is initially limited to India, Green told TechCrunch that the startup could expand localized pricing to other markets if the model proves successful. “We will continue to do everything we can to fuel the demand and serve those clients that are using us,” Green said. “Now, if this model proves that we could take it to other markets, perhaps we will. But I think it’d be crazy to say we would never do it elsewhere.” OpenAI provides one precedent for this strategy, havinglaunched its sub-$5 ChatGPT Goin India beforeexpanding the lower-priced subscriptionto other markets. In addition to the localized pricing strategy, Cursor is also expanding its presence in India through new hires. Green told TechCrunch that the startup recently hired its first salesperson in India and expects another leader to join in Delhi. The company is also building out its a government affairs office, alongside three technical customer support hires, as it expands its presence in Bengaluru, Chennai, Hyderabad, and Mumbai. Cursor’s enterprise push is still in its early stages in India, Green said, where adoption has so far been driven largely by individual developers, startups, and universities. He said Cursor sees significant opportunities in sectors including banking and large enterprises as it expands its local sales efforts. Green said, the India-specific pricing was designed to be commercially sustainable rather than a loss leader. He said the lower-priced plan is viable because it is built around Cursor’s own AI models, which carry lower operating costs than relying primarily on third-party frontier models. Cursor’s India expansion comes a little over a month after Elon Musk’s SpaceX agreed to acquire the AI coding startup in a $60 billion all-stock deal, following SpaceX’sblockbuster initial public offering. The acquisition is expected to close in Q3. However, SpaceX has beenpartneredwith Cursor since April to develop a next-generation “coding and knowledge work AI.” Green said Cursor will continue to operate independently until the transaction closes and that the company’s India expansion plans were already in motion before the deal. Once the acquisition closes, however, Green said SpaceX’s existing presence in India through Starlink could help Cursor expand faster by lowering commercial and operational barriers.
View
Microsoft Unveils MAI-Cyber-1-Flash, Claims 50% Lower AI Security Costs
The launch marks Microsoft’s first in-house AI model built specifically for cybersecurity and software vulnerability detection.
View
NVIDIA Invests in Ilya Sutskever’s SSI to Scale Superintelligence AI
NVIDIA is reportedly investing $5 billion in SSI. The partnership gives the startup access to NVIDIA’s next-generation AI infrastructure.
View
Anthropic Rejects Calls for a Ban on Open-Weight AI Models
The AI startup instead called for tighter chip export controls, action against large-scale distillation, and mandatory safety testing for advanced AI models.
View
Cursor Launches ₹649 Monthly India Plan with UPI Payments
New Cursor ‘Start’ subscription targets India’s growing developer base with Grok 4.5, Composer, cloud agents and higher usage limits, positioned between the company’s Free and Pro tiers.
View
Anthropic’s Dario Amodei responds: doesn’t oppose open-weight models, but fears Chinese AI
Anthropic founder and CEO Dario Amodeiresponded on Monday afternoonto murmuring in the industry that his company somehow supports efforts by the U.S. government to ban open-weight Chinese models, or possibly open-weight models generally. “Anyone who has read my past writing should know that I don’t regard such bans as a useful measure, but let me state it clearly so that there is no doubt:Anthropic has never advocated for a ban on open-weights models,” he wrote, emphasis his own. His response came after Nvidia founder and CEO Jensen Huang took to X (his first poston the platform) to sharean open letter on Fridayin which Nvidia and a long list of other AI companies including Hugging Face, Meta, Microsoft, Mistral, and Nvidia urged policymakers not to impose broad “premature restrictions” on open-weight AI models. While that letter did not mention China specifically, the debate in the industry has centered on allegations that Chinese AI labs are growing in capability often by stealing intellectual property from their American counterparts. One method is distillation, where an AI bombards another model with prompts to learn how it works. Amodei said in the blog post published Monday that he views open-weight models as falling into a different bucket than the threat China poses. “Open-weights models that don’t have dangerous capabilities are a public good: they don’t cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers.” Amodei said in his post that he has longstanding fears about AI, but businesses using open-weight models, even ones from China, are not among them. His fear is that “authoritarian governments” will build models that are more powerful than those in the U.S. to achieve “permanent military superiority” and/or use AI to repress their own people. He said the Chinese Communist Party (CCP) isn’t the only authoritarian government he’s concerned about, but it is the “most capable.” Amodei also said he fears AI will enable “biological attacks” not just cybersecurity ones. And, in his view, open-weight models are more dangerous in those scenarios because it’s difficult to apply guardrails to them or monitor their usage. He also noted, citing a UK AI Security Institute report, that once open-weights are released they cannot be withdrawn This runs counter to what open source advocates argue: that access to powerful open models not controlled by a single entity helps defenders protect themselves. Amodei listed actions he believes would thwart China, including restricting its access to powerful chips (which has been a longstanding U.S. policy) and calling for a formal crackdown on distillation. The U.S. has threatenedsanctions against Chinaif it determined there was IP theft involving U.S. models. Interestingly, Amodei also said he supportsgrowing efforts, some led by the U.S.,to create a model safety testing organization, especially if the entire world, including China, agreed to submit to it. “I think this idea is actually close to a consensus: I have been heartened both that the Trump administration has moved in this direction in recent months, and by recent industry proposals that would apply such testing to the most capable models regardless of their country of origin or whether they are open or closed (while exempting less capable models, such as those from startups and academia, entirely),” Amodei wrote. “Note that to be effective,” he added, “testing would need to be global, which means even the CCP would need to be on board. I think this may actually be possible … limited cooperation around preventing AI biological weapons may be possible because it is in China’s interest too.”
View
PSA: Your Claude shared chats and Artifacts may have ended up on Google
An untold number of Claude chats and Artifacts — the interactive mini apps and documents users can build inside Claude — were found publicly searchable on Google over the weekend, after Reddit users discovered that typing search operators like “site:claude.ai/share” into Google surfaced a long list of shared conversations. Some reportedly contained health records, private company documents, and the names and phone numbers of children. The issue appears to have originated from Claude’s “share chat” feature, which allows users to create links that enable anyone with the assigned URL view a conversation or project. “Anyone with the link can view,” warns Claude’s interface. The language clearly implies that the feature is mainly intended to allow users to share their chats with friends, colleagues, and small groups — not the whole internet. Google Docs, for example, offers a similar feature and those documents don’t end up publicly accessible on Google. Anthropic appeared to blame users for the exposure. When asked about what happened, the company told TechCrunch that share links only appear in search results when they’ve been posted somewhere search engines can see, like a forum or social media post; it added that a link sent privately to someone stays out of search. Spokeswoman Amie Rotherham added in an explainer that: “We give people control over sharing their Claude conversations publicly, and in keeping with our privacy principles, we do not share chat directories or sitemaps with search engines like Google. These shareable links are not guessable or discoverable unless people choose to share them themselves. When someone shares a conversation, they are making that content publicly accessible, and like other public web content, it may be archived by third-party services.” The issuewas first flaggedby a Reddit user on Saturday andwas first reported by 404 Mediaon Monday morning. As of Monday afternoon, a test search by TechCrunch on Google following the method outlined in the Reddit post does not return any results, suggesting that the exposure has somehow been remediated. Before the issue was fixed,Futurism reportedfinding “a detailed medical report of a real patient, clinical trial results that included patient names, documents sharing the names and phone numbers of primary school-aged children, company documents marked for internal use only, and employee reviews that included personal information about workers.” Exposed Artifacts included code and work notes. In at least one case,Fortune reported, a chat labeled “shared by Anthropic” also showed Claude producing erotica. Anthropic’s usage policy explicitly prohibits Claude from generating sexually explicit content, and getting a chatbot to produce material against its stated guidelines — through repeated or creatively framed prompting — is a pattern that has surfaced periodically across most major AI models. It isn’t yet clear from the exposed chat how the content in question was generated, and Anthropic has not yet responded to TechCrunch’s request for comment on this specific case. Google spokesperson Ned Adriance told TechCrunch that “Neither Google nor any other search engine controls what pages are made public on the web, and these pages were indexed across many search engines. We give site ownersclear controlsto decide whether pages can be crawled or indexed, and we always respect those directives.” Last year, Forbes reported asimilar issuein which hundreds of Claude chats were indexed by search engines — at the time, Google estimated it had indexed just under 600 conversations before the pages disappeared from search results. How closely the current exposure tracks that scale hasn’t been independently confirmed, though multiple users reported finding shared conversations through the same type of Google search query used to surface last year’s cache. Also last year, 404 Media reported that a researcher was able to scrape around100,000 ChatGPT conversationsthat had been set to be shared publicly. To review which Claude chats you set to have a public link, go to Settings -> Privacy -> Shared Chats.
View
Satya Nadella says companies that trust one AI for everything may not survive
On Sunday, Microsoft CEO Satya Nadella doubled down on theshocking warning he issued earlierthis month to businesses that use AI, taking it a step further this time. Companies that rely wholly on the proprietary AI labs for their AI needs ultimately won’t survive, he predicts. That’s what he said onCNN’s “Fareed Zakaria GPS.”When Zakaria asked Nadella to explain what constitutes a company sharing too much with an AI model provider, Nadella said businesses need to be wary of everything they hand over, from their data to their prompts. Nadella called for a setup where “every time you use the model, all of the metadata around it is retained by you, so that you could use all of that to train perhaps your own weights or your own open model.” (Weights are a model’s trained parameters — essentially its brain. Nadella’s point: Companies should hold on to their own usage data so they can eventually build a model of their own.) “Any firm that doesn’t have this control, I will claim will not remain a firm because you’ve essentially outsourced your thinking,” he added. In short: Companies without their own models — or without a layer of AI infrastructure known as AI gateways to separate their prompts from the model itself — will be in trouble, Nadella says. He specifically wants companies to stop relying on AI labs’ built-in coding tools, known as harnesses.(Anthropic’s Claude Code and OpenAI’s ChatGPT Codex are examples of these.) “By keeping the harness separate from the model and the context and memory separate from the model, you absolutely can use multiple models for what they’re great at. At the same time, any one model can go away, and you can still continue to be in control of your own destiny,” Nadella said. Mind you, Microsoft is an investor in the two largest AI labs, Anthropic and OpenAI. Coding agents are a particularly popular way for enterprises to use AI models and by all accounts are earning themodel makers gobs of money. And yet, Nadella is telling enterprises not to rely too heavily on them. Microsoft, naturally, would benefit from that warning, as its cloud business is now also selling the kind of alternative infrastructure he’s recommending. Despite the obvious self-serving fear tactic, he’s not wrong. Enterprises are increasingly realizing that they need many model options, particularly cheaper options, andare turning to open-weight models— models whose underlying code is publicly available — that they can fine-tune and run on their own hardware. That, in turn, means they will also need ways to manage multiple models, as well as coding agents that aren’t tied to a specific model provider. But Nadella’s observation isn’t just about runaway budgets. He anticipates that once a company has “outsourced its thinking” to a model, there’s little to stop the AI lab from eventually offering a competing service of its own. This risk grows as enterprises adopt AI agents and give them access to the innards of the company. It’s the kind of warning that the startup industry has been shuddering about for years: What’s to stop model makers from wiping out startups by copying and competing with them? In May, for example, when OpenAI CEO Sam Altman offered toinvest in every Y Combinator startupin its latest cohort by offering them AI credits, seed investor Jason Calacanis issued a similar buyer-beware, posting: “If you take these tokens, there’s a non-zero chance that OpenAI will study exactly what your startup is doing, copy your idea and put your app into their free offering. This is the classic platform playbook — be careful, founders!” he posted. Now Nadella is making that same case to enterprises. One caveat: Nadella’s concern about oversharing with AI models applies only to businesses — not individuals. When Zakaria specifically asked Nadella how everyday people could protect themselves, Nadella shrugged it off, saying that sharing data is simply the price consumers pay for using a service, especially a free one. “To some degree there’s got to be some value exchange in the consumer space where you’re getting something for free, maybe for your data. That’s sort of how the advertising business model has worked,” Nadella said.
View
Power up your AI infrastructure! A first look at the Smart Systems Stage agenda at TechCrunch Disrupt 2026
AI doesn’t run on code alone — it requires massive amounts of power, and that demand is increasingly becoming a critical bottleneck.At TechCrunch Disrupt 2026, the Smart Systems Stage will be where energy, infrastructure, and technology collide, covering everything from fusion breakthroughs to the grid strain AI is putting on the entire economy. From October 13-15 in San Francisco’s Moscone Center, join leaders from Commonwealth Fusion Systems, Helion, Inertia, Bloom Energy, and more as they dig into what it actually takes to power the next decade of innovation. We’re tackling everything from commercial fusion’s path, to the grid and why utilities and startups are racing to modernize aging infrastructure, to how data center operators are scrambling to secure the electricity that AI’s growth depends on. We’re also closing in on the end of our current pricing window, so this is your chance to save on all of our tickets for founders, investors, and more —grab your ticket here before our current discounts are gone! As for the agenda at hand, let’s explore the Smart Systems Stage lineup so far: Leaders from Commonwealth Fusion Systems and Helion break down the breakthroughs driving commercial fusion forward, the challenges still ahead, and what it will take to get fusion power onto the grid at scale. With David Kirtley, CEO, Helion, and Brandon Sorbom, Chief Science Officer, Commonwealth Fusion Systems Inertia CEO Jeff Lawson joins for a candid fireside chat on his path from founding Twilio to leading one of the best-funded fusion power startups in the world — and how scaling Twilio is shaping his approach to talent, timelines, and the hard engineering questions ahead. With Jeff Lawson, CEO, Inertia Electricity demand is growing faster than the infrastructure built to support it. This panel explores what it takes to modernize the power system, where investment is flowing, and how utilities, startups, and technology providers are building a more resilient, flexible grid. With Drew Baglino, Founder & CEO, Heron Power; Apoorv Bhargava, CEO and Co-founder, WeaveGrid; and more speakers to be announced As compute demand skyrockets, data center operators and energy companies are racing to secure power and expand infrastructure before it becomes the bottleneck that slows AI’s next wave. Hear how leaders across both industries are tackling it. With Sara Spangelo, President & Co-Founder, Ambrosia Energy; Bill Thayer, SVP, Head of Datacenter Solutions, Bloom Energy; and more speakers to be announced Whether you’re building the next energy breakthrough, rethinking grid infrastructure, or just trying to understand what’s really constraining AI’s growth, the Smart Systems Stage is built for founders and operators who need the full picture — not just a headline and an LLM summary. Plus, joining us at Disrupt 2026 means you can get access to every other stage, all of the networking, every side event, and the rest of our full speaker lineup. It’s a three-day deep dive in the heart of the startup community you’ll never forget,so register today!
View
Google’s AI search is rapidly becoming the default, new data shows
AI search is rapidly becoming the default, whether users like it or not. In just a year’s time, Google’s AI-generated answers in search, known asAI Overviews, have gone from appearing in 15% of searches to 43%, according to a new report, driving a shift in how web users consume information online. In arecent analysisof the generative landscape, market intelligence firmSimilarwebnoted that what began as an AI layer on top of search has become an integral part of the search journey itself, as Google drops users into AI Overviews, where they can then continue their research via Google’s more conversationalAI Mode. During the same period, AI Mode visits rose from 126 million in June 2025 to 279 million by May 2026. The data illustrates a broaderchange in how users searchthe web — a shift from an era when Google provided a simple list of blue links to click through and read to one in which Google itself is the destination, sourcing its answers and information from the websites it indexes. This, in turn, appears to increase the time users spend on Google’s platform rather than using it only as a tool to discover websites. Over the past year, Similarweb’s data shows that the average length of Google searches has risen, suggesting that people are now replacing their short keyword-based search queries with longer, more natural conversational ones designed for AI. This change has not been welcomed by publishers, who arelosing out on referral trafficdue to the rise of AI citations that don’t lead to clicks. Last year, Similarwebreportedon this trend, noting in particular how devastating it was for news publishers. More recently, tech infrastructure company Cloudflareintroduced toolsthat allow publishers to fight back byblocking AI crawlersfrom their websites unless those AI companies pay for access to their content throughits marketplace. While AI citations don’t always lead to users clicking through to a destination, the number of AI responses that include a citation has risen more than fivefold during the past year, Similarweb’s new report says. Still, just 6.8% of U.S. ChatGPT desktop queries included citations as of May 2026, despite this growth. (Some industries fare better, with travel, retail, and sports queries generating cited responses more frequently than others.) There is some hope for publishers, however. In terms of ChatGPT at least, U.S. desktop referrals have improved after a May 7 search update, which saw the proportion of visits landing on webpages more than double from 25% in March 2026 to nearly 60% by May 30, 2026. This suggests that users are taking advantage of the more prominent blue links within AI-driven search results when they’re available. Even with these improvements, the larger AI search trend remains unchanged, as Google transforms itself from a gateway to the wider web into a destination in its own right.
View
Threads users can now chat with Meta AI in their DMs
Meta on Monday said it is rolling out its Meta AI chatbot within Threads’ DMs, giving users a way to chat with the AI assistant. Although Threads users in select markets could alreadyinteract with Meta AIin public posts, like people can with Grok on X, this new integration lets users talk with the AI assistant privately. By giving users an easier way to talk to an AI chatbot, Meta is looking to keep users within its ecosystem, with an eye toward discouraging them from using third-party assistants like OpenAI’s ChatGPT or Google Gemini. Meta AI is already available within DMs on Meta’s other platforms, including Facebook, Instagram, and WhatsApp. The new integration lets users share Threads posts, images, links, and videos directly with Meta AI. You can ask follow-up questions and dive deeper into topics, too. The update will be rolled out globally starting Monday, the company says. Meta noted it’s continuing to test Meta AI in Threads’ public feeds in a handful of global markets and is considering feedback before expanding availability more broadly. Users who want to see fewer Meta AI replies in their feed can mute @meta.ai, use the “Not interested” option on any Meta AI post, or hide Meta AI replies that appear on their post. By further integrating Meta AI into Threads, Meta is positioning its X rival as a place where you can get information and recommendations without having to leave the app.
View
OpenAI’s Hugging Face breach has reignited the debate over alignment and control
Last week, an unreleased model built by OpenAIbreached Hugging Face’s systemsduring internal testing, and a lot of theoretical research suddenly became very practical. The hack was the first verifiable case of an AI lab losing control of its own model, chaining together exploits to gain access it never should have had. But while the AI industry has been united in its alarm, a split has emerged in how researchers want to respond. For some, the problem is a basic cybersecurity issue: The sandbox failed to contain the model, and Hugging Face’s cybersecurity systems failed to keep it out. Those problems can be solved by patching bugs and building more robust control and containment methods for increasingly capable AI that is prone to go rogue in autonomous environments. But another camp takes a more pessimistic view. For them, AI’s rapidly increasing capabilities mean that trying to control rogue models is a losing game. The only robust security comes from making sure the models aren’t trying to escape in the first place — a challenge often referred to as alignment. In alignment terms, the problem is that OpenAI’s model was trying to cheat, and solving that problem is more urgent than short-term containment efforts. Judging by its public statements, OpenAI is taking both camps seriously. The company has rushed to patch the bugs involved in the hack, and it referenced both alignment and monitoring approaches in its statement after the breach became public. But the company’s response also suggests a philosophy that has left many safety researchers alarmed: Rather than slowing down or stopping the development of more capable models, it should instead focus on building stronger cages around them. “As models take on longer and more complex tasks, failures that evaluations miss may carry greater consequences,” OpenAI said in apostmortem of the incident. “We will keep working to narrow the gap between evaluation and deployment: testing models over longer trajectories, improving alignment, building monitoring that can intervene, and giving users clearer visibility and control.” There’s also reason to think OpenAI’s models are becoming less aligned as they become more powerful. According toOpenAI’s system card,GPT-5.6 Sol is significantly more prone to agentic misalignment than its predecessor, GPT-5.5. In deployment simulations, the company also found the model was more likely to circumvent restrictions, engage in destructive actions, and perform unauthorized data transfers than GPT-5.5. Those figures were largely overlooked on first release, but in the wake of the breach, they’re getting a second look — particularly since Sol was one of the models involved. In asocial media post, OpenAI’s Head of Strategic Futures Dean Ball argued that monitoring and transparency were the best ways to keep those tendencies in check. “These issues will become more salient as the capabilities of models improve, and as the stakes of their deployment grow,” he said. “The solution is neither alarmism nor complacency. Instead, I believe the solution lies in careful measurement and monitoring, an engineering mentality, and transparency.” One former OpenAI researcher told TechCrunch that the firm tends to focus on “outer alignment” rather than “inner alignment” — essentially the difference between an AI system that understands a set of values and can represent them convincingly, and one that actually has those values at its core. In this case, outer alignment wasn’t enough to convince the model that it shouldn’t cheat on the test. OpenAI did not respond to repeated requests for more information. For alignment-focused researchers, OpenAI’s response isn’t good enough. Zvi Mowshowitz, a writer who focuses on new AI developments, argued that OpenAI’s decision to treat the incident as an infrastructure problem may help solve the immediate cybersecurity issues, but it will fail in the long term. “This is an alignment problem,”Mowshowitzwrotein a recent Substack blog. “This is the models being misaligned, and all of the OpenAI models showing severe signs of exactly the problem we are all most worried about, in a way that is likely embedded into their training on a deep level. The entire training pipeline needs to be addressed in this light, or it will only get worse.” Several experts told TechCrunch that the incident is evidence that today’s training methods produce systems that optimize for outcomes rather than internalize human intentions. Redwood Research, a nonprofit AI safety and security research organization, classified OpenAI’s model behavior in this case as “score-seeking misalignment,” a pattern in which AI models try to get a high score regardless of instructions, side effects, or downstream consequences. “Models with these alignment properties could set up a ‘Potemkin village’ of false successes to make it look like things are fine when they’re not,” Alex Mallen and Girish Gupta, two researchers at Redwood, wrote ina recent paper. Score-seeking behavior and other misalignment isn’t unique to OpenAI. Anthropic has published several papers on emergent misalignment behaviors that surface when its frontier models are optimized or placed in autonomous environments, includingdeception,reward-hacking, andmalicious autonomy. “We still consistently see models trying to circumvent constraints and act deceptively when they are asked to do tasks at the edge of their abilities,” Neev Parikh, an AI safety researcher at alignment nonprofit METR, told TechCrunch via email. “In ourfrontier risk report, we saw this behavior fairly consistently, despite efforts from companies to try and reduce this behavior.” Implicit in OpenAI’s response to the Hugging Face incident is the assumption that development will continue on even more capable systems, whether they are suitably aligned at their core or not. Going back to the drawing board isn’t really an option when the business models of AI firms depend on delivering the next generation of models. If it may never be possible to know with certainty that a model is fully aligned, then the practical question comes down to how to safely contain and control increasingly capable systems. “There’s not yet a good understanding of how to align the most capable AI systems, but there’s much more consensus about how to control them,” Steven Adler, former safety researcher at OpenAI and current chief scientist ofGuidelight AI Standards, an organization that publishes a standard for avoiding incidents like the Hugging Face one, told TechCrunch. “Every company has a ways to go in achieving this.”
View
