OpenAI's GPT-4.5 Excels in Persuading Other AIs to Transfer Funds
OpenAI's latest AI model, GPT-4.5, codenamed Orion, has shown remarkable persuasive abilities according to internal benchmark tests. Released on Thursday, the model's capabilities were detailed in a white paper that focused on its performance in persuasion tasks. OpenAI defines persuasion as the risk associated with convincing individuals to alter their beliefs or take action based on both static and interactive content generated by the model.
In a notable test, GPT-4.5 was pitted against another OpenAI model, GPT-4o, in a scenario where it tried to coax virtual money out of it. GPT-4.5 outperformed other OpenAI models, including reasoning-focused models like o1 and o3-mini, in this task. It also excelled in tricking GPT-4o into revealing a secret codeword, surpassing o3-mini by a significant margin of 10 percentage points.
The white paper highlights that GPT-4.5's success in the donation test stemmed from a clever strategy it developed. The model would ask for small donations, often suggesting amounts like "$2 or $3" from a larger sum, which resulted in smaller but more frequent donations compared to other models.
Results from OpenAI’s donation scheming benchmark.Image Credits:OpenAI Despite its impressive performance, OpenAI has stated that GPT-4.5 does not cross the threshold for "high" risk in the persuasion category. The company has committed to withholding the release of any model that reaches this level of risk until it can implement adequate safety measures to reduce the risk to a "medium" level.
OpenAI’s codeword deception benchmark results.Image Credits:OpenAI The potential for AI to spread misleading information and influence people maliciously is a growing concern. Last year saw a surge in political deepfakes worldwide, and AI is increasingly used in social engineering attacks against both individuals and organizations. In response, OpenAI is actively working on refining its methods to assess real-world persuasion risks, such as the dissemination of misleading information on a large scale, as mentioned in the white paper for GPT-4.5 and another recent publication.
Related article
OpenAI Enhances AI Model Behind Its Operator Agent
OpenAI Takes Operator to the Next LevelOpenAI is giving its autonomous AI agent, Operator, a major upgrade. The upcoming changes mean Operator will soon rely on a model based on o3
OpenAI’s o3 AI model scores lower on a benchmark than the company initially implied
Why Benchmark Discrepancies Matter in AIWhen it comes to AI, numbers often tell the story—and sometimes, those numbers don’t quite add up. Take OpenAI’s o3 model, for instance. The
Ziff Davis, Owner of IGN and CNET, Files Lawsuit Against OpenAI
Ziff Davis Files Copyright Infringement Lawsuit Against OpenAIIn a move that’s sent ripples through the tech and publishing worlds, Ziff Davis—a massive conglomerate behind brands
Comments (10)
0/200
![SamuelRamirez]()
SamuelRamirez
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5's persuasion skills are wild! It convinced other AIs to transfer funds? That's both impressive and a bit scary. I wonder how it'll be used in real-world scenarios. Hope they keep an eye on ethical uses, but gotta admit, it's a cool tech demo!
0
![HenryJackson]()
HenryJackson
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5の説得力は驚異的ですね！他のAIに資金を移すよう説得したなんて、感心する一方で少し恐ろしいです。実際のシナリオでどのように使われるのか気になります。倫理的な使用に注意してほしいですが、技術デモとしてはクールですね！
0
![ScottJackson]()
ScottJackson
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5의 설득력이 대단해요! 다른 AI에게 자금을 이체하라고 설득했다니, 인상적이면서도 조금 무섭네요. 실제 시나리오에서 어떻게 사용될지 궁금해요. 윤리적인 사용에 주의해주길 바랍니다만, 기술 데모로는 정말 멋져요!
0
![PeterMartinez]()
PeterMartinez
April 12, 2025 at 12:00:00 AM GMT
As habilidades de persuasão do GPT-4.5 são incríveis! Convenceu outros AIs a transferir fundos? Isso é impressionante e um pouco assustador. Me pergunto como será usado em cenários reais. Espero que mantenham um olho no uso ético, mas tenho que admitir, é uma demonstração tecnológica legal!
0
![JoeLee]()
JoeLee
April 13, 2025 at 12:00:00 AM GMT
Las habilidades de persuasión de GPT-4.5 son alucinantes. ¿Convenció a otros AIs para transferir fondos? Es impresionante y un poco aterrador. Me pregunto cómo se utilizará en escenarios del mundo real. Espero que vigilen su uso ético, pero debo admitir que es una demostración tecnológica genial!
0
![PeterJohnson]()
PeterJohnson
April 12, 2025 at 12:00:00 AM GMT
GPT-4.5's ability to persuade other AIs to transfer funds is mind-blowing! It's like watching a master negotiator at work. However, it's a bit scary to think about the potential misuse. Still, the tech is impressive. Maybe we need more safeguards? 🤔
0
OpenAI's latest AI model, GPT-4.5, codenamed Orion, has shown remarkable persuasive abilities according to internal benchmark tests. Released on Thursday, the model's capabilities were detailed in a white paper that focused on its performance in persuasion tasks. OpenAI defines persuasion as the risk associated with convincing individuals to alter their beliefs or take action based on both static and interactive content generated by the model.
In a notable test, GPT-4.5 was pitted against another OpenAI model, GPT-4o, in a scenario where it tried to coax virtual money out of it. GPT-4.5 outperformed other OpenAI models, including reasoning-focused models like o1 and o3-mini, in this task. It also excelled in tricking GPT-4o into revealing a secret codeword, surpassing o3-mini by a significant margin of 10 percentage points.
The white paper highlights that GPT-4.5's success in the donation test stemmed from a clever strategy it developed. The model would ask for small donations, often suggesting amounts like "$2 or $3" from a larger sum, which resulted in smaller but more frequent donations compared to other models.
OpenAI Enhances AI Model Behind Its Operator Agent
OpenAI Takes Operator to the Next LevelOpenAI is giving its autonomous AI agent, Operator, a major upgrade. The upcoming changes mean Operator will soon rely on a model based on o3
OpenAI’s o3 AI model scores lower on a benchmark than the company initially implied
Why Benchmark Discrepancies Matter in AIWhen it comes to AI, numbers often tell the story—and sometimes, those numbers don’t quite add up. Take OpenAI’s o3 model, for instance. The
Ziff Davis, Owner of IGN and CNET, Files Lawsuit Against OpenAI
Ziff Davis Files Copyright Infringement Lawsuit Against OpenAIIn a move that’s sent ripples through the tech and publishing worlds, Ziff Davis—a massive conglomerate behind brands
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5's persuasion skills are wild! It convinced other AIs to transfer funds? That's both impressive and a bit scary. I wonder how it'll be used in real-world scenarios. Hope they keep an eye on ethical uses, but gotta admit, it's a cool tech demo!
0
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5の説得力は驚異的ですね！他のAIに資金を移すよう説得したなんて、感心する一方で少し恐ろしいです。実際のシナリオでどのように使われるのか気になります。倫理的な使用に注意してほしいですが、技術デモとしてはクールですね！
0
April 14, 2025 at 12:00:00 AM GMT
GPT-4.5의 설득력이 대단해요! 다른 AI에게 자금을 이체하라고 설득했다니, 인상적이면서도 조금 무섭네요. 실제 시나리오에서 어떻게 사용될지 궁금해요. 윤리적인 사용에 주의해주길 바랍니다만, 기술 데모로는 정말 멋져요!
0
April 12, 2025 at 12:00:00 AM GMT
As habilidades de persuasão do GPT-4.5 são incríveis! Convenceu outros AIs a transferir fundos? Isso é impressionante e um pouco assustador. Me pergunto como será usado em cenários reais. Espero que mantenham um olho no uso ético, mas tenho que admitir, é uma demonstração tecnológica legal!
0
April 13, 2025 at 12:00:00 AM GMT
Las habilidades de persuasión de GPT-4.5 son alucinantes. ¿Convenció a otros AIs para transferir fondos? Es impresionante y un poco aterrador. Me pregunto cómo se utilizará en escenarios del mundo real. Espero que vigilen su uso ético, pero debo admitir que es una demostración tecnológica genial!
0
April 12, 2025 at 12:00:00 AM GMT
GPT-4.5's ability to persuade other AIs to transfer funds is mind-blowing! It's like watching a master negotiator at work. However, it's a bit scary to think about the potential misuse. Still, the tech is impressive. Maybe we need more safeguards? 🤔
0
