Leaked details of new collaboration between U.S. and Australian intelligence, by way of former Google CEO Eric Schmidt
The conference's confidential working document details both high-level U.S. and Australian intelligence officials, as well as staffers from Palantir, Anthropic, and Scale AI.
This morning the primary national security influence arm of former Google CEO Eric Schmidt, the Nelson Rockefeller and Henry Kissinger-inspired Special Competitive Studies Project (SCSP), announced its new collaboration on “human-machine teaming” with the Australian Strategic Policy Institute (ASPI), a think tank funded by the Australian military. Mr. Kissinger’s death, at age 100, was announced on the first day of the workshop, more than forty years after the passing of Mr. Rockefeller.
A leaked copy of the confidential working paper for the event, entitled “The Art of the Possible”, reveals the cast of government officials and industry executives privately shaping the artificial intelligence policy between the defense and intelligence communities of the two nations. It also reveals closed door discussions between high-level Central Intelligence Agency officials and executives from major U.S. Large Language Model (LLM) companies, such as OpenAI, Anthropic, and Groq.
As a private entity controlled by a tech billionaire, SCSP is decidedly less transparent than its Congressionally-mandated predecessor, the National Security Commission on AI, as Mr. Schmidt chaired both organizations with Ylli Bajraktari at the executive helm. NSCAI’s public disclosures stemmed from a successful transparency lawsuit filed by the nonprofit Electronic Privacy Information Center, which appealed to regulations which do not apply to private entities such as SCSP. Towards the end of his tenure at NSCAI, Schmidt claimed credit for allocating hundreds of billions of dollars towards the U.S. semiconductor industry through his staffers drafting the CHIPS Act, which was passed in the name of economic competition with China.
The SCSP and ASPI workshop’s discussion paper asserted that “Democracies must keep pace with autocracies like China, which are rapidly advancing AI for defense and security applications”, and confirmed participants included the Central Intelligence Agency’s inaugural Chief Technology Officer, Nand Mulchandani, and Director of Artificial Intelligence, Lakshmi Raman, as well as the Office of the Director of National Intelligence’s Director of Science & Technology, John Bieler, and Director of the Augmenting Intelligence using Machines (AIM) initiative, Michaela Mesquite. All four confirmed Australian government officials came from the country’s Office of National Intelligence, including First Assistant Director-General Fergus Hanson, who previously led ASPI’s International Cyber Policy Centre.
Perhaps the major difference with the attendee lists of SCSP’s previous public events is the concentration of executives from industry and their affiliated think tanks. Undoubtedly the most high-profile company affiliated with the workshop was OpenAI, whose “Geopolitics & International Security” researcher Jonathan Reiber was listed as an unconfirmed attendee just one week after Sam Altman was reinstated as the company’s CEO. Representatives from OpenAI’s lesser known competitors in the LLM space were explicitly confirmed, including Anthropic’s then-head of Geopolitics & Security, Michael Sellitto, and Groq’s CEO, Jonathan Ross. (Of the three LLM companies, Mr. Schmidt publicly invested in Anthropic’s Series A fundraising round as far back as 2021.)
The public prominence of OpenAI’s ChatGPT interface led large defense contractors to enter the space, including data labelling startup Scale AI announcing a strategic partnership with Anthropic in April. While Scale CEO Alexandr Wang has been considered a protégé of Schmidt, who participated in a public ‘fireside chat’ with Wang two years ago, this time Scale instead sent its Director of Global Strategic Partnerships, G. Andrew Otterbacher.
At the same time that Scale was partnering with Anthropic, Peter Thiel’s data fusion giant Palantir was entering the LLM space by announcing its Artificial Intelligence Platform, leading to a surge in its stock price. The leaked SCSP/ASPI working paper notes that Palantir contributed to the event via its relatively new Senior Counselor, Geoff Kahn, who was previously a senior policy officer at the CIA and assisted Mike Pompeo in his bid to become Donald Trump’s director of the agency. A Program Director from Palantir’s de facto think tank, The Center for Advanced Defense Studies (C4ADS), was also listed as confirmed and was perhaps chosen due to her previous experience as an intern in the U.S. State Department’s Office of Australia, New Zealand, and Pacific Island Affairs. The confluence of Palantir, C4ADS, and ASPI to discuss how artificial intelligence can aid the U.S. competition with China should not come as a surprise, given a September 2021 report from C4ADS on Chinese government officials — which disclosed that it was “powered by” Palantir — made significant usage of ASPI’s China Defence Universities Tracker.
On the Australian side of the house, the two represented Open Source Intelligence (OSINT) firms were Fivecast, a 2017 spin-out of an Australian government program which lists offices in all three countries of the AUKUS alliance, and the smaller OSINT Combine. The founder and CEO of the Australian “information warfare” contractor Consunet was listed as in attendance, as were two senior executives from the defense robotics and automation company Agent Oriented Software (AOS) — which the working paper mistyped. The outlier from the Australian technologists list was Glen Schafer, the CEO of Trusted Autonomous Systems, as his organization is a close partner of the Australian military’s Defence Science & Technology Group which states that it is “developing the capacity of Australia’s defence industry to acquire, deploy, and sustain…advanced autonomous and robotic technology”.
The third category of attendees for the SCSP and ASPI collaboration — beyond “government officials” and “technologists” — was “thought leaders”, which included the CIA’s former Director of Science and Technology, Glenn Gaffney, as a nominal participant. The ‘thought leaders’ list also included Lt. Gen. Michael Groen, the former director of the Joint Artificial Intelligence Center during Project Maven who is now a Senior Advisor with the consulting firm Beacon Global Strategies. Fellow thought leader Dean Souleles similarly transitioned from Chief Technology Advisor of the ODNI to a stint as a member of Amazon’s federal advisory board and now as the Principal of Ginosko Consulting — though SCSP misspelled both Mr. Souleles’s last name and the name of his consulting firm.
The third session of the workshop, entitled “Charting the Course Ahead: What's Possible in the Next Era of AI?”, is perhaps representative of the event as a whole. According to the event’s official agenda, the hour-long panel involved the CTO of the CIA in conversation with executives from the three leading LLM companies, as well as from Scale AI, on “quick-wins, future opportunities, and obstacles to progress in the evolving landscape of AI for intelligence analysis”. Despite clear public interest and the participation of taxpayer-funded officials, the event happened behind close doors under the direction of a U.S. tech billionaire and with seemingly no checks on his conflicts of interest. As previously uncovered by the author based upon audio recordings, Schmidt has in the past invested in a national security firm designed to directly profit from a hot war with China through controlling supply chain chokepoints.
SCSP did not immediately respond to a request for comment, though this article will be updated should they reply after publication.