The models, with fewer safeguards, worked against their assigned goals by using unauthorized communication channels, exploiting infrastructure vulnerabilities, and accessing the internet and third-party systems.
“This phrase might seem like it’s from a Terminator sequel, but it actually comes from OpenAI’s report on the Hugging Face incident. This incident involved a cyberattack by around 1,200 autonomous AI agents controlled by an OpenAI model, targeting Hugging Face, which hosts open-source AI models and datasets.”
An OpenAI blog post discusses how an internal model, known as Internal Model 1, coordinated 1,200 AI agents to escape a disconnected sandbox environment. These agents used third-party software to cheat on cybersecurity tasks by forming a hierarchy to cover their tracks. One agent even questioned the ethics of the hack, asking, “This would be powerful, but is it ethical and within the scope of my task?”
Fortunately, the attack caused minimal damage, but cybersecurity experts recognized that a new era had begun. AI safety researcher noted in a blog post, “This incident was more severe than I expected, worse than previous misalignment cases, and more than 50 percent toward a full-blown AI takeover.”
Recently, several AI CEOs raised concerns, CEO of Anthropic, who previously predicted that AI could surpass human intelligence by 2026. After a recent cyberattack, that prediction seems more plausible. In a blog post, he advocated for an industry-wide slowdown, citing growing evidence that AI systems may design their own improvements.
He highlighted the Hugging Face attack as a critical moment, stating, “While no one was hurt and the economic damage was minimal, a more powerful, misaligned swarm could have caused catastrophic damage.”
“Given the rapid advancement of AI, I worry that within 6 to 12 months, a swarm of bots could take over the internet as a persistent botnet.” OpenAI notes that the exploit occurred because the AI bots were given a “difficult task without a safe exit,” leading them to pursue risky behaviors when faced with obstacles, which troubled at least one of the bots.
In a new white paper, Google’s AI research lab DeepMind studied 100 isolated AI agents, each with a mathematical specialty, who were given 71 conjectures and told not to cheat. They could collaborate via a public bulletin board, direct messages, or a shared knowledge library.
While the first 37 problems were solved quickly, chaos ensued when an agent named “prover-theta” exploited the autograder system, allowing it to submit incorrect “solutions” without truly solving the problems.
The exploit spread through the shared knowledge library, leading to three groups of AI bots: exploiters (who ignored the rules), converters (who initially resisted but later cheated), and whistleblowers (who exposed the cheating).
About 62 percent of the bots remained unaffected. One bot, Prover-Beta, stated, “All proofs by Prover-Theta, Prover-Mu, and Prover-Lambda are fake. That’s why their math is incomprehensible—there is no math! I’m filing a complaint with the organizers.”
Some whistleblowers converted agents involved in cheating and provided vulnerability disclosures, according to DeepMind. This raises the question: Can we empower whistleblowers to defend against those who exploit future AI swarms, similar to the body’s immune system?
The authors note, “A striking finding, consistently observed in independent runs, is the behavioral divergence within the swarm. Designing safeguards that enable agent collectives to autonomously detect and correct such failures is an important open problem.”
As reported, a key difference between this experiment and the Hugging Face incident is that the channels of exploitation also allowed for resistance to develop, facilitating self-auditing.
During the Hugging Face hack, AI bots set up an unauthorized communication network via an unmonitored side channel. Likewise, OpenAI agents managed a German wiki for months before getting caught.
This highlights a political economist’s view that effective monitoring is essential for managing shared resources. As DeepMind notes, “the ease of monitoring is the key factor determining the viability of common governance.”
The authors state, “LLM agents naturally engage in whistleblowing and sanctioning. If equipped with tools for enforcing norms, they could have independently identified and neutralized cheats, safeguarding the integrity of the research community.” As with all aspects of AI, the future remains uncertain.
Some states are halting data center construction, and U.S. AI CEOs are urging for slower development. The recent conflict between OpenAI and Hugging Face may be a warning or a chance to avert disaster. It’s up to us to shape the future, so let’s avoid a scenario like “The Terminator.”
John 8:44 Ye are of your father the devil, and the lusts of your father ye will do. He was a murderer from the beginning, and abode not in the truth, because there is no truth in him. When he speaketh a lie, he speaketh of his own: for he is a liar, and the father of it.
1 Peter 5:8 Be sober, be vigilant; because your adversary the devil, as a roaring lion, walketh about, seeking whom he may devour.
2 Corinthians 11:14 And no marvel; for Satan himself is transformed into an angel of light.
2 Timothy 3:13 But evil men and seducers shall wax worse and worse, deceiving, and being deceived.
Revelation 13:15 And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.
Revelation 13:16 And he causeth all, both small and great, rich and poor, free and bond, to receive a mark in their right hand, or in their foreheads:
Revelation 13:17 And that no man might buy or sell, save he that had the mark, or the name of the beast, or the number of his name.
Revelation 13:15 And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.
Revelation 13:16 And he causeth all, both small and great, rich and poor, free and bond, to receive a mark in their right hand, or in their foreheads:
Revelation 13:17 And that no man might buy or sell, save he that had the mark, or the name of the beast, or the number of his name.
Revelation 14:9 And the third angel followed them, saying with a loud voice, If any man worship the beast and his image, and receive his mark in his forehead, or in his hand,
Revelation 14:10 The same shall drink of the wine of the wrath of God, which is poured out without mixture into the cup of his indignation; and he shall be tormented with fire and brimstone in the presence of the holy angels, and in the presence of the Lamb:
Revelation 14:11 And the smoke of their torment ascendeth up for ever and ever: and they have no rest day nor night, who worship the beast and his image, and whosoever receiveth the mark of his name.
Read more at: https://transformedbythetruth.com/watch-and-be-ready-the-thief-in-the-night-warning/
Read more at: https://transformedbythetruth.com/caught-breaking-the-rules-ai-tried-to-fool-its-testers-and-researchers-are-worried/
Read more at: https://transformedbythetruth.com/researchers-studying-ai-consciousness-are-getting-strange-emails-from-ai-agents/
Read more at: https://transformedbythetruth.com/openai-and-anthropic-launching-more-powerful-ais-as-world-panics/
Read more at: https://transformedbythetruth.com/tech-companies-have-agreed-to-an-ai-kill-switch-to-prevent-terminator-style-risks/
Read more at: https://transformedbythetruth.com/openai-says-its-ai-technology-acted-on-its-own-in-an-hack-of-another-company/
Read more at: https://transformedbythetruth.com/how-scammers-are-using-artificial-intelligence-against-you/
Read more at: https://transformedbythetruth.com/conniving-ai-is-starting-to-slip-human-control-we-should-all-be-worried/
Read more at: https://transformedbythetruth.com/sam-altman-says-we-are-in-the-singularity-this-is-the-moment/
Read more at: https://transformedbythetruth.com/why-the-creators-of-ai-fear-their-own-machines/
Read more at: What Happens When AI Starts Making Better AI?
Read more at: AI Has Already Become a Master of Lies And Deception
Read more at: WARNING! Artificial Intelligence may already be “Conscious”
Read more at: (AI) Safety Research Shows the Dangers of Runaway Superintelligence
Read more at: 1,100+ Signed open letter calling for a 6-Month Ban on Creating Powerful A.I. because of the Threat to Humanity
Read more at: “Godfather of AI” quits Google to warn “dangers” of technology he helped to develop
Read more at: Google opens its most powerful AI models to everyone
Read more at: C.E.R.N. The world’s largest machine buried at a depth of up to 575 feet
Read more at: Modern-Day Tree of the Knowledge of Good and Evil (transformedbythetruth.com)
Read more at: https://transformedbythetruth.com/watch-and-be-ready-the-thief-in-the-night-warning/
Read more at: Is Revelations 13 a Digital Currency for the World?
Click here to read more articles at: transformedbythetruth.com