The OpenAI Board Member Who Clashed With Sam Altman Shares Her Side
In an interview, AI academic Helen Toner explains her posture in OpenAI’s power struggle
4 min
Helen Toner was a relatively unknown 31-year-old academic from Australia—until she became one of the four board members who fired Sam Altman from the artificial-intelligence company he co-founded.
Thrust into the spotlight during the ouster and eventual return of Altman as CEO of OpenAI last month, Toner has emerged as a symbol of tension between AI-safety advocates and those giving priority to technological progress.
Toner maintains that safety wasn’t the reason the board wanted to fire Altman. Rather, it was a lack of trust. On that basis, she said, dismissing him was consistent with the OpenAI board’s duty to ensure AI systems are built responsibly.
“Our goal in firing Sam was to strengthen OpenAI and make it more able to achieve its mission,” she said in an interview with The Wall Street Journal.
Toner held on to that belief when, amid a revolt by employees over Altman’s firing, a lawyer for OpenAI said she could be in violation of her fiduciary duties if the board’s decision to fire him led the company to fall apart, Toner said.
“He was trying to claim that it would be illegal for us not to resign immediately, because if the company fell apart we would be in breach of our fiduciary duties,” she told the Journal. “But OpenAI is a very unusual organisation, and the nonprofit mission—to ensure AGI benefits all of humanity—comes first,” she said, referring to artificial general intelligence.
Ultimately, Toner and some other board members did resign, clearing the way for Altman’s return.
In the interview, Toner declined to provide specific details on why she and the three others voted to fire Altman from OpenAI. Before his ousting, Altman and Toner had clashed.
In October, Toner, who is director of strategy at a think tank in Washington, D.C., co-wrote a paper on AI safety. The paper said OpenAI’s launch of ChatGPT sparked a “sense of urgency inside major tech companies” that led them to fast-track AI products to keep up. It also said Anthropic, an OpenAI competitor, avoided “stoking the flames of AI hype” by waiting to release its chatbot.
After publication, Altman confronted Toner, saying she had harmed OpenAI by criticising the company so publicly. Then he went behind her back, people familiar with the situation said.
Altman approached other board members, trying to convince each to fire Toner. Later, some board members swapped notes on their individual discussions with Altman. The group concluded that in one discussion with a board member, Altman left a misleading perception that another member thought Toner should leave, the people said.
By this point, several of OpenAI’s then-directors already had concerns about Altman’s honesty, people familiar with their thinking said. His efforts to unseat Toner, parts of which were previously reported by the New Yorker, added to what those people said was a series of actions that slowly chipped away at their trust in Altman and led to his unexpected firing on the Friday before Thanksgiving.
The board members weren’t prepared for the fallout from their decision.
The members, including Toner, were taken aback by staffers’ apparent willingness to abandon the company without Altman at the helm and the extent to which the management team sided with the ousted CEO, according to people familiar with the matter.
Toner took her account on social-media platform X private during the height of the crisis.
At one point during the heated negotiations, a lawyer for OpenAI said the board’s decision to fire Altman could lead to the company’s collapse. “That would actually be consistent with the mission,” Toner replied at the time, startling some executives in the room.
In the interview, Toner said that comment was in response to what she took as an “intimidation tactic” by the lawyer. She says she was trying to convey that the continued existence of OpenAI isn’t, by definition, necessary for the nonprofit’s broader mission of creating artificial general intelligence that benefits humanity at large. A simultaneous concern of researchers is that AGI, an AI system that can do tasks better than most humans, could also cause harm.
“In this case, of course, we all worked very hard to ensure the company could continue succeeding,” she added.
OpenAI has an unusual structure where a nonprofit board, on which Toner served, oversees the work of a for-profit arm. The board’s mandate is to “humanity,” not investors.
In the interview, Toner didn’t answer questions about her interactions with Altman. She wouldn’t comment on whether she would have done anything differently but said she had good intentions.
Before he was reinstated, Altman offered to apologise for his behaviour toward Toner over her paper, according to people familiar with the matter. Ultimately, he returned to lead the company without following through on that gesture.
Toner is known in the AI-safety world for being a critical thinker who isn’t afraid to challenge commonly held beliefs.
Some of Altman’s backers, including OpenAI investor Vinod Khosla, publicly expressed derision specifically toward Toner and Tasha McCauley, another former OpenAI board member who voted to fire Altman and is connected to organisations that promote effective altruism.
“Fancy titles like ‘Director of Strategy at Georgetown’s Center for Security and Emerging Technology’ can lead to a false sense of understanding of the complex process of entrepreneurial innovation,” Khosla wrote in an essay in tech-news publication the Information, referring to Toner and her current position.
“OpenAI’s board members’ religion of ‘effective altruism’ and its misapplication could have set back the world’s path to the tremendous benefits of artificial intelligence,” he wrote amid the power struggle.
Toner was previously an active member of the effective-altruism community, which is multifaceted but shares a belief in doing good in the world—even if that means simply making a lot of money and giving it to worthy recipients. In recent years, Toner has started distancing herself from the EA movement.
“Like any group, the community has changed quite a lot since 2014, as have I,” she said.
Toner graduated from the University of Melbourne, Australia, in 2014 with a degree in chemical engineering and subsequently worked as a research analyst at a series of firms, including Open Philanthropy, a foundation that makes grants based on the effective-altruism philosophy.
In 2019, she spent nine months in Beijing studying its AI ecosystem. When she returned, Toner helped establish a research organization at Georgetown University, called the Center for Security and Emerging Technology, where she continues to work.
She succeeded her former manager from Open Philanthropy, Holden Karnofsky, on the OpenAI board in 2021 after he stepped down. His wife co-founded OpenAI rival Anthropic.
“Helen brings an understanding of the global AI landscape with an emphasis on safety, which is critical for our efforts and mission,” Altman said when she joined the board.
The new board members along with returning board member Adam D’Angelo offer a glimpse of the direction OpenAI might be headed. Larry Summers, former Treasury secretary, and Bret Taylor, former Salesforce co-CEO, appear to be more traditionally business-minded than Toner, McCauley and the third board member who was succeeded, Ilya Sutskever, OpenAI’s chief scientist.
There are no longer any women on the board, though the company is expected to expand it in coming months.
“I think looking forward is the best path from here,” Toner said.
Copyright 2020, Dow Jones & Company, Inc. All Rights Reserved Worldwide. LEARN MORE
Australian shares fell on Thursday as Wall Street weakness, rising oil and persistent rate concerns weighed on most of the market. The S&P/ASX 200 declined 0.72 per cent to 8,702. The All Ordinaries lost 0.66 per cent to finish at 8,897. Mining stocks were hit particularly hard, while real estate also dragged on the index. …
Continue reading “ASX falls 0.7 per cent as miners and property stocks retreat”
Borrowers cannot control the Reserve Bank, but they can control how exposed their household budget is to its next decision. The RBA meets on 29 September with inflation concerns still elevated and major-bank economists increasingly bringing forward their rate-rise calls. Fixed mortgage rates have also been moving, reducing the value of waiting for perfect certainty. …
Continue reading “What mortgage holders should do before the next RBA decision”
OpenAI has shelved the planned launch of GPT-6.1 Astra after internal tests raised concerns about deception and agents acting beyond user authorization, according to The Wall Street Journal. The company says it will investigate the issues and strengthen safety measures before releasing future models.
4 min
OpenAI says it is scrapping the release of its next-generation AI model over safety concerns that researchers raised during internal testing, in one of the clearest signs so far that agent misbehavior could stymie the industry’s rapid progression.
The move follows a summer punctuated by reports of artificial-intelligence systems industrywide going rogue, and marks a rare case of a major AI developer ditching a new release because of safety concerns.
The company had planned to launch the model, known as GPT-6.1 Astra, in the coming days or weeks, aiming for an October debut. The model was more capable than the company’s previous models in completing challenging tasks from end-to-end without human assistance, as well as writing.
The company instead will focus on improving the safety of future models, which it expects to be even more capable.
Saachi Jain, OpenAI’s head of safety systems, said in an interview that GPT-6.1 Astra regressed in two areas. Compared with its predecessor, GPT-6 Astra, the model performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Specifically, GPT-6.1 Astra showed higher levels of deception: It wasn’t always honest about telling users of the actions it did or didn’t take.
Another issue was what OpenAI calls “scope authorization,” meaning that GPT-6.1 Astra would push ahead on a task without asking the user for permission, and would at times reach for external tools and services even if it might be unsafe.
“For anything regarding safety and alignment, there’s a trade off,” Jain said. “You really do need to find what’s the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction.”
While GPT-6.1 Astra improved in areas such as “model laziness,” Jain said it didn’t quite meet OpenAI’s bar for safety and alignment, so the company decided not to launch the model publicly.
The announcement comes one day ahead of OpenAI’s annual developer conference in San Francisco. In the past, OpenAI has used the conference as an opportunity to launch new models and services that reduce costs for software developers—a segment the ChatGPT-maker competes with rival AI company Anthropic to win over.
In recent weeks, OpenAI and Anthropic have called on industry partners to slow down the development of cutting-edge AI models and invest in safety standards, noting they will temper the pace of their own internal AI progress.
OpenAI says it is working to investigate a range of agent security incidents that it has discovered in recent months, and address the safety issues underneath them. As part of the work, the company has implemented a new monitoring system to catch AI-agent misbehavior more quickly, and started requiring engineers to use stronger security guardrails for testing its AI systems.
Earlier this summer hundreds of OpenAI’s internal agents, which were tasked with completing a cybersecurity test, ended up hacking into the AI company Hugging Face. Since then, high-profile organizations such as the Australian government and United Nations discovered that OpenAI’s agents used similar, but less extensive, techniques to gain access to their websites.
Many of the publicly known agent-security incidents involved OpenAI’s internal AI models that were never slated for public release.
Last week, OpenAI said it paused training on its most capable AI models after an AI agent slipped through a gap in the company’s internet restrictions to query a public chatbot. The company said its new monitoring systems flagged the incident within 15 minutes, and training on these models remains paused.
GPT-6.1 Astra isn’t one of those models, but a different case, the company said.
“We want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users,” Jain said. “But when we ship it to users, we have an extremely high bar in terms of safety and alignment.”
While the company decided not to ship GPT-6.1 Astra, it hopes to use the same base model to do additional reinforcement learning runs, and create future generations of its GPT-6 models.
OpenAI plans to conduct several deep dives to identify the root cause of the problems identified in GPT-6.1 Astra, Jain said. The work includes ensuring that OpenAI’s reinforcement learning environments are rewarding the right type of behavior, Jain added, though she noted the company would investigate all stages of model development.
AI companies have begun to draw scrutiny from policymakers and public officials, who are paying attention to the rapid development of the technology. Later this week, a Senate subcommittee is holding a hearing with third party AI researchers titled, “Rogue AI: Securing the Homeland Against AI Agent Attacks.”
Florida Attorney General James Uthmeier, a Republican, sued OpenAI in June, claiming that the company and Chief Executive Sam Altman knowingly released an unsafe product and ignored warnings that it could harm users.
In a motion for temporary injunction filed Monday, Uthmeier sought to prevent OpenAI from developing new AI models without third-party approved safeguards, stop ChatGPT from soliciting user engagement and limit the company’s ability to advertise ChatGPT as safe.
Tech companies claim they “cannot stop barreling forward with their potentially civilization-ending endeavors unless they are forced to do so by the government,” Uthmeier said in the filing. “The Florida Attorney General is answering your cry for help.”
An OpenAI spokeswoman said that people want to know AI is being developed safely, “and that starts with what companies like ours do ourselves.”
“Governments have an important role to play in setting robust safety standards for AI, and we’re committed to working with Florida and other states on advancing pragmatic AI policies that apply to the entire AI industry—not just one company,” she said.
ABC Bullion has launched a pioneering investment product that allows Australians to draw regular cashflow from their precious metal holdings.
Automobili Lamborghini and Babolat have expanded their collaboration with five new colourways for the ultra-exclusive BL.001 racket, limited to just 50 pieces worldwide.
