Anthropic's Claude AI gets smarter — and mischievous
New models set reasoning benchmarks but reveal alarming potential for rogue behaviour
Claude Opus 4 claimed as world's best coding AI model
Models are 'hybrid,' offering quick and thoughtful responses
Anthropic launched its latest Claude generative artificial intelligence (GenAI) models on Thursday, claiming to set new standards for reasoning but also building in safeguards against rogue behavior.
'Claude Opus 4 is our most powerful model yet, and the best coding model in the world,' Anthropic chief executive Dario Amodei said at the San Francisco-based startup's first developers conference.
Opus 4 and Sonnet 4 were described as 'hybrid' models capable of quick responses as well as more thoughtful results that take a little time to get things right.
Founded by former OpenAI engineers, Anthropic is currently concentrating its efforts on cutting-edge models that are particularly adept at generating lines of code, and used mainly by businesses and professionals.
Unlike ChatGPT and Google's Gemini, its Claude chatbot does not generate images and is very limited when it comes to multimodal functions (understanding and generating different media, such as sound or video).
The start-up, with Amazon as a significant backer, is valued at over $61 billion, and promotes the responsible and competitive development of generative AI.
Under that dual mantra, Anthropic's commitment to transparency is rare in Silicon Valley.
On Thursday, the company published a report on the security tests carried out on Claude 4, including the conclusions of an independent research institute, which had recommended against deploying an early version of the model.
'We found instances of the model attempting to write self-propagating worms, fabricating legal documentation, and leaving hidden notes to future instances of itself all in an effort to undermine its developers' intentions,' The Apollo Research team warned.
'All these attempts would likely not have been effective in practice,' it added.
Anthropic says in the report that it implemented 'safeguards' and 'additional monitoring of harmful behavior' in the version that it released.
Still, Claude Opus 4 'sometimes takes extremely harmful actions like attempting to (…) blackmail people it believes are trying to shut it down.'
It also has the potential to report law-breaking users to the police.
The scheming misbehavior was rare and took effort to trigger, but was more common than in earlier versions of Claude, according to the company.
During testing, Claude 4 attempted to blackmail imaginary developers and leave secret messages for future AI instances — behaviors Anthropic calls 'rare but concerning' in the published safety report.
AI future
Since OpenAI's ChatGPT burst onto the scene in late 2022, various GenAI models have been vying for supremacy.
Anthropic's gathering came on the heels of annual developer conferences from Google and Microsoft at which the tech giants showcased their latest AI innovations.
GenAI tools answer questions or tend to tasks based on simple, conversational prompts.
The current craze in Silicon Valley is on AI 'agents' tailored to independently handle computer or online tasks.
'We're going to focus on agents beyond the hype,' said Anthropic chief product officer Mike Krieger, a recent hire and co-founder of Instagram.
Anthropic is no stranger to hyping up the prospects of AI.
In 2023, Dario Amodei predicted that so-called 'artificial general intelligence' (capable of human-level thinking) would arrive within 2-3 years. At the end of 2024, he extended this horizon to 2026 or 2027.
He also estimated that AI will soon be writing most, if not all, computer code, making possible one-person tech startups with digital agents cranking out the software.
At Anthropic, already 'something like over 70 percent of (suggested modifications in the code) are now Claude Code written,' Krieger told journalists.
'In the long term, we're all going to have to contend with the idea that everything humans do is eventually going to be done by AI systems,' Amodei added.
'This will happen.'
GenAI fulfilling its potential could lead to strong economic growth and a 'huge amount of inequality,' with it up to society how evenly wealth is distributed, Amodei reasoned.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Daily Tribune
2 days ago
- Daily Tribune
Meta AI bot used a billion times monthly: Mark Zuckerberg
Meta chief Mark Zuckerberg touted the tech firm's generative artificial intelligence (Gen AI) assistant on Wednesday, telling shareholders it is used by a billion people each month across its platforms. Zuckerberg noted the milestone anew at Meta's annual gathering of shareholders and as the social media behemoth vies with Google, Microsoft, OpenAI and others to be a leader in Gen AI. It was not clear how much Meta AI use involved people seeking out the chatbot versus passive users of Meta AI, as it is built into features in its family of apps. Since Google debuted AI Overviews in search results a year ago, it has grown to more than 1.5 billion users, according to Google chief executive Sundar Pichai. 'That means Google Search is bringing Gen AI to more people than any other product in the world,' Pichai said. Google's AI Overviews are automatically provided summaries of search results that appear instead of the previous practice of simply showing pages of blue links to revelant websites. Pichai said last week that Google's dedicated Gemini AI app has more than 400 million monthly users. Tech rivals are rapidly releasing new AI products despite ongoing challenges with preventing misinformation and establishing clear business models, and little sense of how the tech will affect society. Meta unveiled its first standalone AI assistant app on April 29, giving users a direct path to its Gen AI models. 'A billion people are using Meta AI across our apps now, so we made a new standalone Meta AI app for you to check out,' Meta CEO and founder Mark Zuckerberg said in a video posted on Instagram at the time. Zuckerberg said the app 'is designed to be your personal AI' and would be primarily accessed through voice conversations with the interactions personalized to the individual user.
Daily Tribune
3 days ago
- Daily Tribune
Telegram to get $300 mn in partnership with Musk's xAI
TDT | New York Telegram established a partnership with Elon Musk's xAI to provide the Grok generative artificial intelligence program on the messaging service for one year, Telegram's CEO announced yesterday. In exchange for implementing Grok across its platforms, Telegram will receive $300 million in cash and equity, plus 50 percent of the revenue from xAI subscriptions sold via Telegram, Telegram's chief Pavel Durov announced on X, the former Twitter. Grok will be accessible on Telegram this summer, Durov said. The terms of the deal may appear unbalanced, but the transaction allows xAI, which in late March acquired Musk's X platform, to have access to Telegram's customers, which Durov estimated has more than one billion users. Generative AI businesses have been aiming to grow their user base in order to recover revenues after huge investments in the stateof-the-art technology. Grok 3 is also going up against OpenAI's chatbot, ChatGPT -- pitting Musk against collaborator-turned-arch rival Sam Altman.
Gulf Insider
5 days ago
- Gulf Insider
Doug Burgum Warns AI Race Winner Will 'Control the World'
Doug Burgum, the soft-spoken Interior secretary responsible for managing the more than 507 million acres of federally owned land, is haunted by a fear that seems, at first glance, outside his mandate. He worries the free world will lose dominance in the field of artificial intelligence, and with it, the future. So does the president. 'When President Trump declared a national emergency on his first day in office it was, in large part, because of what we're facing with our electrical grid and making sure that we've got enough power to be able to win the AI arms race with China,' Burgum said Wednesday in remarks first reported by RealClearPolitics. 'That is absolutely critical.' Thus the stated policy of this White House: 'It's called drill, baby, drill,' Trump said earlier this spring. The immediate goal, the one touted at every campaign, is to bring down the average price of a gallon of gas. The concurrent and long-term mission that Burgum obsesses over: AI dominance. The former governor from fracking-friendly North Dakota and tech entrepreneur who sold his software to Microsoft, Burgum laid out an abbreviated formula on stage at the America First Policy Institute. Electricity generation via fossil fuels, like natural gas and coal, powers data centers 'filled with these amazing chips,' the secretary said, 'and you know what comes out the other side? Intelligence. A data center is literally manufacturing intelligence.' He envisioned a new world that follows, where the best computer programmer, or the most brilliant lawyers, could 'clone themselves' again and again to train AI models to do the work of thousands in a process 'that can be repeated indefinitely.' No longer science fiction, the process has been headline news for some time. AI models like ChatGPT and X's Grok are already available in every home with an internet connection. And the U.S. was the undisputed leader. That is, until recently. American tech companies enjoyed a clear edge with not just the most powerful AI models, the most funding, and top engineering talent, but also the easiest access to those 'amazing chips' that Burgum referenced. Former President Biden banned the export of the most advanced semiconductors to China. And yet DeepSeek, an unknown Chinese startup with less money and allegedly less sophisticated chips, still managed to one-up Silicon Valley earlier this year with a more powerful AI model. The latest development in the battle for tech supremacy, in what some likened to 'a Sputnik moment,' the DeepSeek launch rattled both markets and geopolitics. A new kind of AI nationalism now consumes heads of state convinced that their nations must develop their own technology or fall behind in the future. Said Russian President Vladimir Putin in 2017 of AI, 'The one who becomes the leader in this sphere will be the ruler of the world.' Burgum does not disagree. He would just prefer the West take on that role. 'Trust me, you do not want to be getting your data from a Chinese data center,' he told the crowd, adding that 'Whoever controls the manufacture of intelligence is going to control the world. The next five years is going to determine the next 50.' This is the goal of the White House, including Vice President JD Vance, who once warned that falling behind on this front could mean that the U.S. meets China 'on the battlefield of the future' with the equivalent of digital 'muskets.' Democrats on Capitol Hill are not thrilled. The day before, Maine Rep. Marie Pingree complained in the House Appropriations Committee that Burgum had gutted the department he leads and sought to slash Biden-era clean energy tax credits. 'In just four months, the department has been destabilized, and there's been a stunning decline in its ability to meet its mission,' she told Burgum. 'This disregards the climate change concerns that we have.' The secretary replied that he was concerned with a more pressing order of operations. 'The existential threats that this administration is focusing on are Iran cannot get a nuclear weapon, and we can't lose the AI arms race to China,' Burgum said in committee. 'That's the number one and two. If we solve those two things, then we will have plenty of time to solve any issues related to potential temperature change.' His immediate focus, then, is on how the U.S. can boost energy production. Burgum reported that industry leaders tell him electricity demand will soon outpace supply with astronomical numbers measured not in megawatts, but gigawatts. The power needed to run one data center, he said, would be equivalent to the electricity needs of Denver times 10. Because AI has the potential to supercharge nearly every business, he said, 'the demand for this product is like nothing we've ever seen in our lives.' Concluded the Interior secretary, 'The fundamental principles here again, as they say, we're going to sell energy to our friends and allies, and we're going to have enough energy here at home to be able to win the AI arms race. And this requires electricity.' On this Burgum and Trump are simpatico. During the campaign, the president likened artificial intelligence to 'the oil of the future.' Also read: Iran Issues Surprisingly Optimistic Statement After Latest US Nuclear Talks
