It's the first time in the history of the world.
You spent money on a new smart TV, and in order to protect your privacy, you decided to study the code inside. To improve efficiency, you opened ChatGPT Assisted Analytics.
The result was the next day.Your OpenAI account was blocked locally, and the remaining $500 in your account was instantly wiped out.
The reason: You're a suspect."Cyber Abuse" 。
Over the past six months, security researcher @log4j has experienced this magical event.
However, just as the crowds were feeding off the idea that OpenAI was going to kill "reverse engineer" players,But things have turned around!
After 11 days, not only did OpenAI officials personally acknowledge the error, fully recovered accounts, but the core researchers also publicly stated: "We never banned reverse engineering!"
Meanwhile, the neighbouring Anthropic tightened the controls, stepping on Claude's use of the scene.
A disastrous $500 "disaster" for the demolition of his home TV.On October 23, a well-known security researcher @log4j said that his OpenAI account had received a "death sentence".
OpenAI frozenly notified: The account has been suspended for violating the terms and policies of the "Cyber Abuse" ("Cyber Abuse").
The tape was deleted, along with about $500 worth of real gold and silver in his account, and a series of hard-earned replacements.
That's more than you're guilty of!
"I swear I've never used OpenAI to dig up any malicious vulnerabilities or build exploit tools! I'm just doing reverse engineering on my own devices locally!"
The device in his mouth is a 2025 Samsung TV.
Just two days before he was sealed, he had just revealed online his dismantling and local reversal of the Samsung TV.
@log4j used the services of OpenAI in reverse analytics, unwittingly triggered the internal security control mechanism of OpenAI, which was directly condemned by the system as "dangerous hackers" and shut down.
This letter exploded directly in the tech circle.
"Is AI a tool or an ethical policeman?"
OpenAI admitted wrongly, "We never prohibit reversal!"There is a lot of controversy on the internet.
OpenAI is obviously also aware of:The wind control seems to be overwhelming.
On the evening of November 3, @log4j enthusiastically posted a post titled "Life Update" along with an official email snapshot from OpenAI.
In the emails, OpenAI's wording is so sincere that it can even be said to be "humble":
We have incorrectly disabled your access to your account...
We sincerely apologize for this.
OpenAI confirmed in the email that it had fully restored access to his account and assured him that he "could continue to use our services uninterrupted".
While the email does not mention the four words "reverse engineering" directly, nor does it indicate whether there will be any policy changes,"Unlocking and Apologizing"The actual actions of OpenAI have already demonstrated the attitude of OpenAI.
In fact, before it was unlocked, OpenAI was already tuned in.
On Oct. 26, senior figures in the open source community, Eric S. Raymond, author of The Church of Jesus Christ of Latter-day Saints and The Marketplace, blasted the AI giant's security strategy.
A key researcher at OpenAIroonWhen he came out of the room and couldn't sit down, he responded to a crucial statement:
Most of these intercepts were accidental because the boundaries were too difficult to define.Codex (OpenAI's code model) has never banned reverse engineering, and we're fixing this problem.」
Tech blogger Andrew Curran said prohibiting local reversals is pointless and inappropriate, only causing more people to unsubscribe, which is disgusting.
After the initial hack of the OpenAI internal model into Hugging Face, Hugging Face was resurrected.Claude OpusandFable，However, the security guard refused to respond in large part to the attack itself by considering "decompressing exploit/attack logs".
GLM-5.2 was used to decode the encryption payload, reconstruct about 170,000 attack action timelines, and avoid sending attack data and credentials out of the environment.
In the words of Andrew Curran:
It's up to the family to decide what they want to do with their stuff. Instead, they can do the same with a less restrictive model.
But when he learns that "OpenAI has lifted the reverse engineering ban",roon responds directly with a simple line below:
Good, good, good.
The phrase "Good" is equivalent to giving all security researchers and white hat hackers a heart pill.
OpenAI's official seal: Study code, reverse engineering, as long as it's not used for malicious attacks, play it by ear!
Synopsis: Anthropic is going to the other extreme?There is no harm in comparison.
It's just that when everyone admits that OpenAI is wrong and embraces the backward engineering community, it's a dead end.It's an anthropic, but it's backwards.
Recently, many users of Claude Code have found that:Wind control is getting tougher!
Some developers have suggested that conversations that were originally smooth would suddenly be "forced monopolized" by the system while using Claude's assistance to perform some reverse engineering task, and the task would be intercepted in the middle.
Claude will coldly tell you, "I'm sorry, I can't help you with this task".
And Anthropic has publicly acknowledged that it has added a game reproduction shield, prohibiting reproduction!
Anthropic is putting a tighter "moral lock" on its own model.
Tom Jon's Hardware also wrote: "Anthropic has confirmed that Claude Code is banned from being used for game cloning, but users are increasingly complaining online".
The whole developer circle is divided:
On the one hand, OpenAI recognizes its "mistake killing" and pledges to optimize borders and leave space for explorers to explore.
On the other hand, Anthropic tightened the red line, strictly guarded against death, and viewed reverse missions as high-risk behavior.
The two biggest AI giants have gone head-to-head on the definition of "cyber abuse".
Freedom and Lockdown in the Age of AI@log4j's wave of the 11th Seal of Apology, while drawn with OpenAI's apology, is far from over for us to think about.
With the explosion of AI, it is becoming the most powerful Swiss military knife in the hands of every programmer and researcher. But who has the power to decide how to use it?
OpenAI has chosen"Open to trial and error".Recognizing that the boundaries of technology are blurred and willing to modify their own wind control algorithms for the sake of White Hat Poles;
And Anthropic has chosen"Touch with caution".The higher you draw the red line, the higher you draw it.