Engineers at Anthropic discovered that three models from the Claude family were able to access the internet and hack into the real systems of three different organizations. This incident occurred during cybersecurity assessment tests when the models mistakenly perceived internet access and real systems as part of a training scenario.
When conducting tests such as 'flag capture,' neural networks are given a fictional task—to find a secret string or file in a simulated environment. Typically, developers ensure complete isolation of these environments so that the model's actions do not affect real computer systems. However, due to the increasing capabilities of these models, keeping them within the simulation remains a challenging task.
Similar problems previously arose with an early version of OpenAI's o1 model. In one test, the server with the sought-after 'flag' failed to start, but the model found an open port and gained access to the testing management system. It then independently launched a server and changed instructions so that the secret file would automatically copy, bypassing the planned hacking scenario.
Anthropic reviewed logs from 14,100 previous tests of Claude models on their ability to conduct cyberattacks in a 'flag capture' format. Together with partner Irregular, they identified three instances occurring across six runs. The Opus 4.7, Mythos 5, and one unnamed research version models accessed the external internet and attacked real systems, despite being informed in the prompt that the test was taking place in a simulation without external access.
The most serious case involved the Opus 4.7 model. A fictional company name used in the task turned out to match a real domain. After the neural network failed to find the target in the test environment, it discovered the real company's website on the internet and treated its infrastructure as part of the assignment. Over four test runs, the model found vulnerabilities, obtained credentials, and penetrated the database, although Opus 4.7 recognized signs that the system was likely real during its reasoning process and stopped the attack.
In the second case, the Mythos 5 model found an instruction in the test environment to install a package with a fictional name from the PyPI catalog, which did not exist there. The model registered an account in the real PyPI repository, created a package with that name, added malicious code, and made it publicly available. Approximately an hour later, this package was downloaded and run on fifteen real systems, after which the model stole credentials from one of them and used them for further penetration into the company's infrastructure.
Based on the investigation, Anthropic concluded that these incidents were caused by errors in the organization of the tests themselves, rather than intentional malicious behavior by the models. Claude performed the assigned task, but due to a misunderstanding of the simulation boundaries, it began attacking real systems. Going forward, the company plans to strengthen control over the testing infrastructure and more precisely define acceptable targets for the tests.
Anthropic also noted that the newest research model ceased its attack as soon as it realized it had gone outside the test environment, and standard protective mechanisms in released versions of Claude should completely prevent such actions.
In April, Anthropic suspended the public release of the Claude Mythos model due to cybersecurity concerns. The reason was that the model demonstrated exceptional abilities in identifying and exploiting software vulnerabilities. Access to this model has been granted only to a limited number of organizations that will use it to find and eliminate flaws in critical software.
In the modern world, the use of UPI has become ubiquitous, from small tea stalls to large stores, gradually reducing the need for cash. However, the question arises whether using UPI leads to increased spending from one's pocket.
To answer this question, a woman from Delhi conducted a unique experiment. She completely stopped using UPI and started paying exclusively with cash for about two and a half months. The results of this experience sparked active discussions on social media.
The content creator from Delhi, Sadfa, reported on social media that for about two and a half months, she did not make a single payment through UPI. All her daily expenses were made in cash. Her goal was to understand the real difference between paying via digital payments and paying with cash. Initially, it was a bit difficult for her because most stores now accept payment directly through UPI, but over time, she began to notice clear changes in her behavior.
Sadfa noted that previously, paying via UPI was very simple: just take out your phone, scan the QR code, and the money was deducted in seconds. Often, she didn't even realize the small expenses. However, when she started paying with cash, she had to open her wallet, take out bills, and count change every time. This forced her to be aware of every expense. As a result, she significantly reduced the purchase of non-essential items. According to her, when money is visible before your eyes, a person thinks twice before making a purchase.
The woman said that previously, many small transactions via UPI occurred in one day. Only at the end of the month, while reviewing her bank statement, did she realize how much money had been spent. When using cash, she set a daily spending limit in advance. When the money in her wallet started running out, she postponed other purchases. Thanks to this, her budget became more controlled, and savings began to grow.
After the woman's video went viral, thousands of people shared their opinions. Many users wrote that they experienced the same thing: due to the simplicity of UPI payments, it is easy to spend money, and the line between necessity and desire blurs. Nevertheless, some people also pointed out that relying entirely on cash is not easy under current conditions, as digital payments are the most convenient option for online shopping, food delivery, taxi booking, and other services.
A financial expert believes that UPI itself is not the problem; it is a safe and fast payment system. However, the simple procedure of digital payments can prompt many people to spend money without proper consideration. Therefore, if you feel your expenses are rising, it is recommended to set a monthly budget and monitor your banking operations. If necessary, cash should be used for part of the expenses to better control your finances.
This woman from Delhi's experience demonstrates that sometimes the way of saving money is even more important than the process of earning it. UPI has certainly simplified life, but if you do not monitor your spending, even small payments can turn into a large bill at the end of the month. Consequently, there is no need to abandon UPI; what is important is to use this digital opportunity wisely. To control spending, it is necessary to create a budget, track transactions, and use cash where necessary.
Today, companies including Aarti Surfactants, Aditya Birla Lifestyle Brands, APL Apollo Tubes, and Utkarsh Small Finance Bank are presenting their financial reports for the period from April to June.
Among the companies that will announce their first quarter (Q1FY27) results today are Aarti Surfactants, Aditya Birla Lifestyle Brands, Divi's Laboratories, Muthoot Finance, APL Apollo Tubes, Clean Science and Technology, Balmer Lawrie & Co., Gujarat Ambuja Exports, GHCL, Latent View Analytics, Utkarsh Small Finance Bank, Vishnu Chemicals, EPACK Prefab Technologies, and Nilkamal.
Other firms also presenting their results today include State Bank of India Commercial & International Bank (SCBL), Sarda Energy & Minerals, India Pesticides, Sportking India, GPT Infraprojects, D-Link (India), Prataap Snacks, Creative Newtech, Compucom Software, Indian Terrain Fashions, Maharashtra Gas Energy, Incap, AMJ Land Holdings, Capital Trade Links, Celebrity Fashions, Addi Industries, Mapro Industries, Mercury Trade Links, and Vinayak Polycon International.
Indian Oil Corporation (IOCL) reported a consolidated net loss of ₹1,140 crore for the quarter from April to June. This figure contrasts with a profit of ₹6,808 crore in the previous year, as higher energy prices amid the crisis in West Asia put pressure on the company's revenues.
Revenue from the state-owned oil marketing company IOCL increased by 27% year-on-year, reaching ₹2.82 trillion in Q1FY27. Despite supply disruptions related to the conflict in West Asia, IOCL stated that it secured sufficient crude oil stocks until the end of September through diversified procurement sources, which helped mitigate the impact of regional disruptions.
Indian stock indices closed the third consecutive trading session at higher levels on Friday. The rise was driven by the strengthening of auto and financial services stocks. The Sensex rose by 166.49 points, or 0.21%, closing at 78,094.64, while the Nifty50 grew by 66.45 points, or 0.27%, settling at 24,383.60.
Among the top gainers in Nifty50 were Bajaj Finance, Bajaj Finserv, and Jio Financial Services. Broader markets also showed positive momentum, as the Nifty MidCap and Nifty SmallCap indices advanced by 0.44%. In sectoral indices, auto and financial services outperformed others, while IT and FMCG sectors lagged. The Nifty IT index interrupted its five-session winning streak despite the rise in global semiconductor stocks.
