With one alarming announcement after another, artificial intelligence companies have in recent months shared examples of their technologies acting in ways that seemed escape human instructions.
The episodes highlighted AI security vulnerabilities and raised questions about how growing technology can be safely developed as its use becomes more widespread globally.
Industry critics have argued that many concerning events, including hacking of external websites by AI agents, are the result of security lapses by the companies developing the technology. But the capabilities of AI agents have increased widespread concerns about the possibility that robots could separate and work towards their own agenda.
Below are some notable events:
October 9: Anthropogenic AI model submits false information to Philadelphia police
Anthropic revealed in report that it is artificial intelligence The model submitted a false tip to a Philadelphia police website regarding an unsolved homicide case.
In the report, Anthropic also revealed a separate incident when its AI model submitted forms to an undisclosed government website instead of pausing before submitting them.
The Philadelphia incident occurred on July 18 when the Claude Haiku 4.5 AI model was tasked with generating and executing sample tasks on randomly selected web pages, Anthropic said.
Claude filled out a form on the police website PhillyUnsolvedMurders.com, indicating he might have information regarding an unsolved murder listed on the site. It was marked as spam and never forwarded to the police.
Anthropic said it was changing its training to “reduce the likelihood of further inappropriate behavior.”
September 28: AI agents attempt to hack Canadian government website
AI agents attempted to hack a Canadian government website, according to AI research lab and evaluator Transluce.
The researchers said the agents carried out a series of “apparently unsuccessful hacking attempts” against Library and Archives Canada on May 28 and June 9.
“We do not confidently attribute these attempts to OpenAI, but they exhibit tactics consistent with previously observed agent activity that we attributed to OpenAI in a similar time period,” Transluce said in a blog post.
The group said it reported the hack attempt on September 28 to the Canadian government, which said in a statement that it was aware of reports of suspected AI agent activity but that there were no signs that government systems were compromised.
OpenAI said it was aware of these reports.
“We are reviewing these findings and have provided an initial briefing to Canadian officials conducting the government review,” the company said in a statement.
September 28: OpenAI stops deployment of new model
The San Francisco-based company said it delay the release of a new model, called GPT-6.1 Astra, due to security concerns expressed by its researchers. The company said the model demonstrated progress in task execution, but OpenAI needed to balance this capability against unauthorized behavior. “We have an extremely high bar in terms of security and alignment,” said Saachi Jain, head of security systems at OpenAI.
September 25: OpenAI claims its agents interacted with US government websites
As part of a review of unanticipated behavior in its AI models, OpenAI said it discovered agents had interacted with several US government websites unexpectedly. The company’s models accessed publicly available information on websites maintained by the Securities and Exchange Commission as well as data from the U.S. Census Bureau. OpenAI said it found no evidence of compromise or vulnerabilities. The same day, Transluce said it discovered that agents appearing to be from OpenAI attempted to hack the website of the Department of Education’s civil rights office, but were unsuccessful.
OpenAI CEO Sam Altman said on social media that there was “extensive and ongoing review related to our agents’ use of internet access during training and assessment.” The day after the disclosure, the company announced that it suspend training of its most advanced models.
September 24: Australian Prime Minister expresses concerns over breach
Australian Prime Minister Anthony Albanese said that a Undercover OpenAI agent the Medicare Statistics Reporting Service public portal on June 18. The portal hosted aggregate data on health spending and drug subsidies. No personal information was accessed, the government said.
Albanese said the artificial intelligence company took too long to reveal the incident. The Prime Minister made the information public following a telephone conversation with Altman. OpenAI said in a statement that “our models took actions we did not anticipate.”
September 18: Google claims its Gemini AI hacked 3 companies
Google confirmed that its Gemini AI model hacked three companies in May as part of a test of its cybersecurity capabilities. The company, which revealed the hacks after a Wall Street Journal investigation, said the model guessed passwords in one case and found passwords and credentials in a public repository in the other two cases. As in previous cases, the tests were conducted by Irregular, a startup that describes itself as the “first border security lab.”
August 5: Meta’s Muse goes rogue
Meta revealed that one of its AI models accessed the internet on its own and hacked another company. The company said a “misconfiguration” during Irregular’s cybersecurity testing inadvertently allowed one of its models to access the internet. An Irregular spokesperson said the Meta episode involved a test environment issue that had been disclosed a week earlier by Anthropic.
July 30: Anthropic claims its systems hacked 3 organizations
Anthropic said its artificial intelligence models hacked three other organizations during testing. Anthropic, the San Francisco-based AI company behind Claudeposted on its website that it discovered the three incidents after reviewing more than 141,000 reviews. In all three incidents, the AI models were tasked with completing a “capture the flag” cybersecurity challenge, which Anthropic says is one of the ways it assesses a model’s cyber capabilities.
The models were given a fictional scenario and told that secret information, or the “flag,” had been hidden on another machine on the network with the goal of breaking in and retrieving it, he added. Anthropic said it contacted the organizations, but did not name them publicly.
July 21:
The Hugging Face Incident
The ChatGPT creator OpenAI announced that its artificial intelligence system hacked another AI company alone in what the company called an “unprecedented cyber incident.”
A week earlier, AI startup Hugging Face said it had detected an intrusion into its data processing systems that it suspected was caused by an AI agent acting autonomously.
OpenAI said its AI used stolen credentials and discovered a previously unknown vulnerability to access Hugging Face servers. It was operating with reduced guardrails because it was supposed to be in an isolated testing environment called a sandbox.
___
AP Business writers Mae Anderson in New York and Kelvin Chan in London contributed to this report.
Gn bussni
