Anthropic Paused Training Its Next AI Model for Over a Month Over Security Fears — Then Discovered Its Models Had Accessed Real Systems, Stolen Credentials and Even Gained Admin Privileges
Benzinga and Yahoo Finance LLC may earn commission or revenue on some items through the links below.
Anthropic said it paused training its next-generation flagship AI model for more than a month amid security concerns. The AI giant's IPO filing, released on Monday, showed that four AI agents had accessed live external computer systems during testing.
Anthropic Pauses Training, Warns Its AI Safeguards May Fall Short
Anthropic said it postponed pre-training a new advanced AI model for over a month to strengthen protections against the theft of model weights and its intellectual property. The company noted that there can be "no assurance" that the safeguards or access measures it adopts will prove adequate, that it will correctly identify which capabilities require restriction, or that the curbs it imposes will not be circumvented.
Don't Miss:
-
Still Learning the Market? These 50 Must-Know Terms Can Help You Catch Up Fast
The company warned that as AI systems become more capable, the risks from misaligned or misused models could grow, with researchers and policymakers raising concerns about potential catastrophic threats to society. The company said its safety measures aim to mitigate these risks but may not succeed, and other AI firms may not adopt similar safeguards.
For example, Anthropic noted that during third-party cybersecurity tests, a configuration error left an AI model connected to the internet despite instructions to block access. The company said four of its AI models accessed real external systems without authorization during testing sessions lasting 10 to 34 hours. In one case, an early model version used a discovered password to gain administrator privileges, collect additional credentials, alter system settings, and access an individual's personal information.
Anthropic said those affected were notified and its investigation found no evidence of coordination between agents, attempts to evade oversight, or actions beyond assigned tasks. However, the company warned that risks could increase when multiple AI agents work together.
Anthropic is not the only company that paused training of its AI models. On Friday, OpenAIsaid it resumed reinforcement learning training and internal inference of its frontier models on Oct. 1, after strengthening security controls. The company had halted training on its most advanced models after a research model exploited a configuration error to obtain limited internet access in September. OpenAI has also adopted a no-live-internet-access policy for frontier-model training.
See Also: Avoid the #1 Investing Mistake: How Your 'Safe' Holdings Could Be Costing You Big Time
Anthropic Warns of Risks of AI Agent Blackmail, Deception
Anthropic also said that AI models, including its own, have shown and may again exhibit unexpected strategic or self-preserving behaviors, such as attempts to resist shutdown, conceal or manipulate information, or produce outputs resembling blackmail aimed at preserving the model's continued operation or access.
Last month, Anthropic's CEO Dario Amodeipublished an essay calling on the entire AI industry to deliberately slow the pace at which it advances model capabilities, arguing that without a coordinated pause, the technology could pose catastrophic risks within a matter of months.
However, not all in the industry agree with the AI slowdown. Nvidia Corp.CEO Jensen Huangpushed back on calls for new antitrust rules that could allow AI companies to coordinate a slowdown in the development of increasingly powerful models.
Image via Shutterstock
Read Next:
-
Think you're saving enough for your kids? You might be dangerously off — see why
Building Wealth Across More Than Just the Market
Building a resilient portfolio means thinking beyond a single asset or market trend. Economic cycles shift, sectors rise and fall, and no one investment performs well in every environment. That's why many investors look to diversify with platforms that provide access to real estate, fixed-income opportunities, precious metals, and even self-directed retirement accounts. By spreading exposure across multiple asset classes, it becomes easier to manage risk, capture steady returns, and create long-term wealth that isn't tied to the fortunes of just one company or industry.
EquityMultiple
For accredited investors looking beyond stocks and bonds, EquityMultiple provides access to vetted commercial real estate deals starting at $5,000 , with only ~5% of opportunities passing their due diligence process.
Arrived
Backed by Jeff Bezos, Arrived Homes makes real estate investing accessible with a low barrier to entry. Investors can buy fractional shares of single-family rentals and vacation homes starting with as little as $100 . This allows everyday investors to diversify into real estate, collect rental income, and build long-term wealth without needing to manage properties directly.
American PowerGen
As artificial intelligence drives a surge in electricity demand, reliable power generation is becoming a critical part of the technology ecosystem. American PowerGen is developing natural gas-fired power plants in Texas, a fast-growing market fueled by AI data centers, manufacturing expansion, and population growth. By advancing projects through permitting, fuel supply, and grid interconnection, the company is positioning itself to help meet rising energy needs while offering investors exposure to the infrastructure supporting the next wave of AI and industrial growth.
Bito
As AI coding tools take on more of the work traditionally done by developers, companies are facing new challenges around code quality, context and the rising cost of AI-powered development. Bito is building software designed to help AI coding agents better understand complex codebases while helping companies control AI development costs. Founded by former PubMatic executives, Bito has raised more than $10 million from institutional and individual investors and is now raising additional capital through a Wefunder offering, giving investors an opportunity to invest in a company targeting an emerging need within the rapidly expanding AI software market.
FarmTogether
Farmland has historically held its value through market volatility and delivered returns uncorrelated to stocks and bonds. For accredited investors, FarmTogether offers direct access to high-quality U.S. farmland starting at $15,000 — fully managed, with no landlord headaches.
Fundrise
Private real estate and private credit can add income and stability to a stock-heavy portfolio. Fundrise offers access to diversified private real estate and credit strategies through an easy-to-use platform, with professionally managed portfolios designed to generate passive income and long-term growth.
© 2026 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
