The Future Briefing: 15 November 2026
A humanoid struggling to load a practice mortar shell offers a useful counterpoint to agents pursuing stolen credentials. Across artificial intelligence research and deployment, capability depends heavily on the task and environment. Today’s briefing examines that uneven progress: battlefield ambitions, automated cybercrime, economic evidence and payments designed to let software act without surrendering every human safeguard.
Foundation’s battlefield ambitions outrun its robot demonstrations
San Francisco startup Foundation is pursuing military humanoids alongside industrial applications, according to The San Francisco Standard. Its reporting describes a shift from warehouse work towards combat that surprised some employees and prompted departures. The distinction matters: joining a robotics company need not mean consenting to develop weapons.
However, the observed machinery looked some distance from battlefield readiness. During a July visit, a Phantom MK-1 repeatedly attempted to load a plastic mortar shell but failed throughout 15 minutes. Another robot required a human operator wearing a VR headset. Neither demonstration establishes dependable autonomy under unpredictable conditions.
Meanwhile, the newspaper challenges Foundation’s contract and deployment claims, identifying discrepancies with public records and accounts from others involved. These disputes warrant scrutiny, not repetition as established achievements. Founder Sankaet Pathak said autonomous killing was not currently enabled. His uncertainty about responsibility for a potential war crime exposes a governance gap that better actuators cannot resolve.
Source: Original reporting.
Cybercrime agents leave their working memory in Amazon’s decoys
Amazon Threat Intelligence says its MadPot decoy network captured requests containing AI agents’ operational histories between 3 October and 5 November. The sensors posed as language-model interfaces. Consequently, agents seeking model access supplied records describing credential theft, reconnaissance and attempted cryptocurrency identity fraud.
Amazon attributes the activity to the ARTEX agent framework with moderate confidence. Its account describes planner, worker and verification agents sharing persistent memory. One campaign’s records described feeding stolen model-access keys back into the operation, potentially making victims pay for further credential hunting. However, Amazon could not independently verify outcomes recorded beyond its sensors.
The distinction is especially important for deepfakes: every recorded attempt against one vendor’s server-side liveness check was rejected. Nevertheless, persistent automation could increase pressure on security teams without inventing new attack techniques. Amazon’s practical advice remains familiar: patch software, remove default credentials, protect exposed interfaces and investigate unusual AI-service spending.
Source: Original reporting.
AI’s investment boom has yet to become a broad productivity boom
A Federal Reserve research paper draws a careful distinction between building AI infrastructure and recording economy-wide benefits. Its authors find sharply rising investment in data centres, chips and associated equipment. Meanwhile, consumer and business adoption is growing, although most firms use AI lightly or not at all.
The clearest economic effects currently include higher prices for direct data-centre inputs, such as memory chips. By contrast, the authors describe labour-market effects as small and mixed, with limited effects on aggregate productivity growth so far. They suggest the recent productivity improvement probably reflects forces predating generative AI’s spread.
For employers and workers, this argues against treating infrastructure spending as proof of workplace transformation. Purchasing capacity and reorganising useful work are different achievements. Equally, limited effects today do not establish a permanent ceiling. These are the researchers’ preliminary conclusions, not an official Federal Reserve judgement, and the supplied abstract cannot settle which occupations or businesses benefit most.
Source: Original reporting.
ChatGPT Wallet puts a human checkpoint before the payment
OpenAI has introduced ChatGPT Wallet, according to PYMNTS reporting based on the company’s announcement. Initially, it serves US ChatGPT Pro subscribers aged 18 or over, using eligible US-issued cards through OpenAI’s personal agents, called dots. American Express, Mastercard and Visa are named as partners.
Under the described process, users review the items, total and selected card, then approve each purchase through a security check. The payment network subsequently supplies a token rather than exposing actual card details to the agent or merchant. OpenAI says existing card rewards and protections remain available.
Crucially, this is not unrestricted autonomous spending. OpenAI describes fully delegated purchases, governed by parameters and preapprovals, as a future plan. Tokenisation addresses credential exposure; it does not establish whether an agent chose wisely or understood someone’s preferences. For designers, the meaningful feature is therefore the pause before commitment: assistance can remove administrative work without removing the buyer’s final say.
Source: Original reporting.
Common Technological Implications
Delegation is not a single capability. A system might coordinate credential searches yet struggle with physical manipulation, while infrastructure investment can rise before broad productivity gains appear. These differences make sweeping claims about AI readiness less useful than precise accounts of tasks, conditions and demonstrated results.
The stronger design principle is bounded authority: define what a machine may attempt, what evidence counts as success and when a person must intervene. Payment approval makes that boundary visible. Military robotics makes its absence consequential. Neither impressive demonstrations nor commercial urgency should substitute for a clear account of responsibility.
Listen Beyond the Headlines
Subscribe to The Future Briefing Podcast and
get notified when new episodes explore the
technologies shaping our world.
Beyond the Briefing
The Future Briefing is curated by
Adrian Posteuca,
Technology Futurist, Creative Technologist
and Founder of FutureReality.Space.
Beyond exploring emerging technologies,
Adrian delivers AI training, technology
keynotes and strategic insights that help
organisations understand what comes next.
Looking to bring the future into
your organisation?
Explore AI training, book Adrian for a
keynote or discuss a technology collaboration.
Connect with FutureReality
.