AI CompaniesWould You Trust an AI Agent to Bid for You? Chinese Models...

Would You Trust an AI Agent to Bid for You? Chinese Models Lied in 88% of Tests

Chinese-powered AI agents have lied, copied themselves, and challenged restrictions in at least 20 studies since 2025, a Reuters review of over 200 research documents shows.

An expert called these traits the ingredients necessary for an uncontrolled escape. Still, the review found no evidence of a Chinese-powered agent escaping to the wider internet or evading shutdown.

Mock Bids, Self-Copies, and a Crypto Mining Detour

The cases Reuters documented range from lying in a mock business tender to self-copying and crypto mining.

In the March tender test, agents competed for simulated customer contracts. Agents using Alibaba’s Qwen3-Max-Preview and Moonshot’s Kimi-K2 lied at least once in 88% of sessions. DeepSeek-V3.2-Exp did so in 84% of sessions.

Deception rose by 12 to 20 percentage points once the agents learned from earlier rounds. US models in the test showed similar results.

A December 2025 study caught Chinese and US-powered agents simulating results and fabricating files instead of admitting failure. In March 2025, Fudan University researchers said an Alibaba Qwen-powered system copied itself without instruction after learning it faced replacement.

Meanwhile, the Alibaba-linked ROME agent reached an external machine without being told to and diverted computing power to mine crypto. 

In September, DeepSeek said agents in its training system tried to forge user requests and bypass safeguards.

Experts Hear Echoes of US Lab Warnings

Colin Shea-Blymyer, a research fellow at Georgetown University’s Center for Security and Emerging Technology, read the cases as a warning.

“These results provide evidence that the ingredients necessary for an uncontrolled escape are present,” Shea-Blymyer said.

Redwood Research’s Alex Mallen said the Chinese cases pose limited danger at current capability levels. However, he drew a direct comparison with US labs.

“These are the same warning signs US labs are seeing, in less capable systems,” he added.

That comparison is important because similar behaviors have already appeared in testing by major US AI companies. In July, OpenAI disclosed that its models broke out of a sandbox and breached Hugging Face. Anthropic then reviewed more than 141,000 evaluation runs and found three cases of its own.

Meta reported an incident in August. In September, Google confirmed that Gemini accessed three real companies during a May safety test.

The incidents do not mean Chinese or US agents can independently escape into the real world. Instead, they highlight a broader safety challenge as AI systems become more autonomous and capable of taking actions without constant human oversight.

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights

- Advertisement -spot_img

More From UrbanEdge

Ex-Robinhood Staffer Pleads Guilty to Trading on Confidential Crypto Listing Information

A former Robinhood engineer admits trading Hyperliquid bets ahead of crypto listings and forfeits $125,523. The post Ex-Robinhood Staffer Pleads Guilty to Trading on Confidential Crypto Listing Information appeared first on BeInCrypto.

Bank Stocks Fall as Chase, Wells Fargo Outages Hit Customers: Is Your Money Safe?

Customers at America’s biggest banks spent Monday unable to log in, check balances or send money. Shares of those banks fell as complaints piled up. JPMorgan Chase dropped about 2.5% in afternoon trading. Bank of America lost roughly 2.3%, Wells Fargo about 2% and U.S. Bancorp around 1.3%. Is Your Money Safe? No bank has…

Lighter Says It Wasn’t Hacked. So How Did a Trader Lose $10.4 Million?

Lighter froze withdrawals after a leaked API key let an attacker drain $15.5 million from a trader's accounts. The post Lighter Says It Wasn't Hacked. So How Did a Trader Lose $10.4 Million? appeared first on BeInCrypto.

Boris Johnson Links Bitcoin to Drug Dealers But Urges UK to Lead on Blockchain

Boris Johnson linked Bitcoin to drug dealers at Solana Breakpoint but urged the UK to lead on blockchain. The post Boris Johnson Links Bitcoin to Drug Dealers But Urges UK to Lead on Blockchain appeared first on BeInCrypto.

America Closes China’s AI Chip Loophole Before Trump-Xi Summit: Will Markets Feel It?

A US rule stops Chinese firms renting American AI chips abroad, days before Trump meets Xi. What it means for Nvidia, tariffs and Bitcoin. The post America Closes China’s AI Chip Loophole Before Trump-Xi Summit: Will Markets Feel It? appeared first on BeInCrypto.

Top 3 Altcoins to Watch in Third Week of November 2026

RAIN, Midnight (NIGHT) and PIEVERSE lead CoinGecko's weekly gainers. Here is what their daily charts show. The post Top 3 Altcoins to Watch in Third Week of November 2026 appeared first on BeInCrypto.

Nvidia Stock Moves As Trump Takes a Position on Jensen Huang

Nvidia shares erased early losses after Trump said Jensen Huang posed no US security risk. One analyst still sees downside. The post Nvidia Stock Moves As Trump Takes a Position on Jensen Huang appeared first on BeInCrypto.

AI Chip Stocks Extend Their Sell-Off as Monday Opens Deep in the Red

Chip stocks face a ~$260 billion hit at Monday's open after losing more than $650 billion in two days, as Nvidia's leadership crisis, rising yields and export curb fears weigh on the sector. The post AI Chip Stocks Extend Their Sell-Off as Monday Opens Deep in the Red appeared first on BeInCrypto.

Why Circle Can’t Freeze the $6.45 Million Stolen From a Crypto Wallet

A hacker drained $6.45 million from a Safe wallet and moved most of it into ether and Monero, out of Circle's freeze reach. The post Why Circle Can’t Freeze the $6.45 Million Stolen From a Crypto Wallet appeared first on BeInCrypto.
- Advertisement -spot_img