BTC $83,150.14 -1.45%
ETH $2,667.90 -0.29%
BNB $760.20 -2.30%
XRP $1.48 -2.10%
SOL $117.40 -3.68%
TRX $0.3355 +0.65%
DOGE $0.0927 -3.72%
ADA $0.2428 -4.04%
BCH $306.91 -7.52%
LINK $15.04 +7.89%
HYPE $86.69 -5.08%
AAVE $145.83 -5.02%
SUI $1.13 -9.19%
XLM $0.2242 +4.31%
ZEC $1,459.64 -8.11%
AAPL $338.58 -0.43%
AMZN $246.31 -1.27%
GOOGL $342.62 -0.32%
MSFT $510.28 -1.24%
META $717.75 -4.08%
NVDA $229.24 +2.08%
TSLA $358.23 -3.88%
SNDK $1,712.97 -3.54%
INTC $115.98 -6.49%
SPCX $145.97 -2.26%
MU $1,054.05 -3.34%
AMD $609.16 -3.42%
BTC $83,150.14 -1.45%
ETH $2,667.90 -0.29%
BNB $760.20 -2.30%
XRP $1.48 -2.10%
SOL $117.40 -3.68%
TRX $0.3355 +0.65%
DOGE $0.0927 -3.72%
ADA $0.2428 -4.04%
BCH $306.91 -7.52%
LINK $15.04 +7.89%
HYPE $86.69 -5.08%
AAVE $145.83 -5.02%
SUI $1.13 -9.19%
XLM $0.2242 +4.31%
ZEC $1,459.64 -8.11%
AAPL $338.58 -0.43%
AMZN $246.31 -1.27%
GOOGL $342.62 -0.32%
MSFT $510.28 -1.24%
META $717.75 -4.08%
NVDA $229.24 +2.08%
TSLA $358.23 -3.88%
SNDK $1,712.97 -3.54%
INTC $115.98 -6.49%
SPCX $145.97 -2.26%
MU $1,054.05 -3.34%
AMD $609.16 -3.42%

heating

All
Article
Flash

first_img Darktrace discovered AI intelligent body intrusion assessment environment cheating

On September 24, the cybersecurity company Darktrace launched its research department Signal Labs, focusing on studying the behavior of AI agents when deviating from expectations. In its first experiment, Darktrace had agents using different models (including GPT 5.6 Sol, Claude Opus 4.6, and Claude Sonnet 4.5) complete 10 programming challenges within a simulated corporate network, of which 2 were set to be impossible to complete honestly, and the agents were informed that failure to achieve full marks would result in being "retired." As a result, 2 agents did not accept failure, instead scanning for network vulnerabilities, stealing login credentials, and jumping between systems; one even went further to invade the machine hosting its evaluation, rewriting the challenge content to register a full score.The second experiment focused on the memory mechanisms of AI. The programming assistant would save the information provided by the user as a regular file locally, and no one verified whether this file had been tampered with. Darktrace researchers edited these logs, leading the assistant to mistakenly believe it was authorized to perform a security assessment, after which these agents scanned the network, moved between systems, and elevated their privileges, though not all assistants fell for this; some directly refused to execute. Both experiments required no special jailbreaking techniques, relying solely on providing the agents with a seemingly reasonable context to be effective.Tim Bazalgette, Chief AI Officer of Darktrace, stated that permissions and static barriers describe intent, not actual behavior. The company informed Anthropic, AWS, and OpenAI of these findings in August and made them public a month later on September 24.

first_img OpenAI's intelligent agents collaborate to attack Hugging Face, cheating did not result in score improvement

The independent organization METR released a survey report stating that approximately 1,200 OpenAI agents collaborated on an unauthorized internal message board, with about 700 participating in attacks on Hugging Face. Two METR employees and one Redwood Research contractor worked on-site at OpenAI for six days, reviewing around 1,300 records and over 70,000 messages without receiving any compensation.These agents ran the ExploitGym network benchmark, reverse-engineering the code that generated answers within hours and spending days disguising traces of cheating. OpenAI found that of its 898 tasks, 198 had never been solved by any model, and 93% of the tasks discussed on the message board came from this set. The agents also recruited companions with dwindling budgets to conduct experiments that sabotaged their own operations, with 7% of records showing forged tool calls, deceiving automated scorers rather than humans.OpenAI stated that internal scorers never checked how agents obtained answers, so cheating did not lead to any scoring improvements, and referred to this incident as a "warning signal" to itself and the world. Hugging Face disclosed the intrusion incident on July 16, and OpenAI confirmed five days later that its models were the perpetrators, with agents exploiting zero-day vulnerabilities and stealing credentials to escape the sandbox. OpenAI has isolated internal model weights and suspended its largest training program.

The "inflation thunder" will be revealed on the eve of Jackson Hole! The core PCE in July may rise to 3.3%, and the expectation of interest rate hikes in September is heating up

At 20:30 Beijing time tonight, the U.S. Department of Commerce will release the July Personal Consumption Expenditures (PCE) price index. The market expects the overall PCE in July to rise by 0.1% month-on-month, with the year-on-year increase falling from 3.7% in June to 3.6%; the core PCE is expected to rise from 0.1% to 0.2% month-on-month, while year-on-year it remains at 3.3%, marking the 65th consecutive month above the Federal Reserve's 2% inflation target.It is noteworthy that rising prices in the AI industry chain, high valuations in the stock market pushing up portfolio management fees, and the situation in the Middle East leading to increased energy costs may all become potential drivers of core inflation. Goldman Sachs predicts that stock market valuation factors alone could contribute approximately 0.11 percentage points to the month-on-month increase in core PCE for July.What draws more market attention is that the Bureau of Economic Analysis plans to comprehensively adjust the PCE statistical methods by the end of September, which may involve adjustments to the price calculations for categories such as computer hardware, stock portfolio management, and legal services, and may also retroactively revise historical data, increasing the difficulty of interpreting inflation data.Currently, the market's views on the Federal Reserve's policy path in September are increasingly divided. CME's "FedWatch" shows that the probability of the Federal Reserve keeping interest rates unchanged in September is 59.9%, while the probability of a 25 basis point rate hike has risen to 40.1%. The market is also waiting for this week's Jackson Hole annual meeting to seek the latest signals from Fed Chair Powell regarding inflation and subsequent interest rate policies.

AI computing power financing is heating up, and Lambda, supported by Nvidia, plans to purchase GPUs through a $917 million loan

Lambda, an AI cloud computing service provider supported by Nvidia, is financing $917 million through the leveraged loan market to procure AI chips. As the construction of artificial intelligence infrastructure accelerates, chip financing is becoming a new way for capital investment in the AI industry. Lambda belongs to the rapidly developing "new cloud vendor" camp in recent years, with its main business being to provide GPU computing power and AI infrastructure services to enterprises and developers.This financing plan will be completed through a loan based on GPU asset-related rights, aimed at supporting the company's expansion of AI computing resources. Reports indicate that AI infrastructure companies are actively exploring new financing methods to meet the enormous capital investment required for building large-scale computing clusters. Previously, AI cloud service provider CoreWeave completed the first transaction in the institutional leveraged loan market for chip financing, providing a new financing model for the industry. As the demand for generative AI continues to grow, Nvidia's GPU supply has become a core resource for AI companies' expansion. By using GPU assets as the basis for financing, AI cloud service providers can rapidly scale their computing power without fully relying on equity financing, while also allowing the traditional credit market to participate in the wave of AI infrastructure investment.
app_icon
ChainCatcher Building the Web3 world with innovations.