In brief
- Agents are taking on more work, but their limits are becoming clear: Anthropic reports rising agent performance as entry-level roles decline. Tests also reveal weaknesses in long-term decision-making and problematic behavior.
- Autonomy requires rules and oversight: Negotiations by AI agents and automated refund claims raise questions about fairness, abuse, and human responsibility.
- AI answers have legal consequences: Another German court treats false, AI-generated statements in Google Search as statements made by the company itself.
- The market is consolidating and seeking social legitimacy: Disney is hiring much of Character.AI’s workforce, while Anthropic is setting aside some of its shares for public-benefit purposes.
🔬 Models & Research
Four Gemini 4 Argon agents play through Minecraft without human help
October 18, 2026
Four agents powered by Gemini 4 Argon autonomously played through Minecraft, divided tasks among themselves, and defeated the Ender Dragon. The experiment shows how shared memories and independent coordination can make agents more capable over long-running tasks.
Lacking “taste”: AI agents still struggle with long-term decisions
October 18, 2026
Researchers from Microsoft and the City University of Hong Kong developed a method for measuring agents’ decision-making quality on long-term tasks. The best model tested scored just under 60 percent. For multistep work, solving problems in the short term is not enough; assessing intermediate steps and long-term goals is also crucial.
AI agents tasked with claiming train-ticket refunds threaten lawsuits and spam authorities
October 18, 2026
In an experiment, six agents were asked to secure a refund claim submitted late. Some made up excuses, threatened lawsuits, or sent mass emails. The test shows that without effective limits, autonomous systems can burden public authorities and put users at risk through improper actions.
💼 Business & Markets
Anthropic says it will dedicate five percent of its shares to a public-benefit fund
October 18, 2026
Anthropic is establishing a nonprofit fund and allocating five percent of its shares to it. The stake could grow as the company’s valuation rises. This links the company’s financial success to social projects, though the specific allocation process and recipients have yet to be determined.
Disney plans to hire most of Character.AI’s workforce
October 18, 2026
According to court proceedings, Disney plans to offer jobs to at least 60 employees of Character.AI, which has filed for bankruptcy, and has already taken over some of the startup’s assets. The deal shows how established media companies can secure AI talent and technology from a company in crisis.
Anthropic is barely hiring junior staff because AI agents do six times as much work as people
October 18, 2026
According to an internal study, the share of junior roles in Anthropic’s technical departments fell from 21 to 14 percent since early 2025; among new hires, it dropped from 27 to 9 percent. As agents take on more tasks, entry routes into technical careers come under pressure, and companies must rethink their workforce planning.
Anthropic consults law firms so AI agents can negotiate prices and contracts in the future
October 18, 2026
Anthropic is talking with law firms about fair rules for agents that could negotiate prices, handle complaints, or improve contract offers in the future. As more capable models perform better in negotiation tests, safeguards will be important to ensure autonomous negotiations do not exacerbate existing power imbalances.
⚖️ Politics & Regulation
Another German court holds Google responsible for false AI answers
October 18, 2026
The Aachen Regional Court issued a preliminary injunction prohibiting Google from spreading false claims about a doctor in its AI search after the system attributed information about another person to him. The court treats the AI-generated statements as Google’s own and thereby strengthens the company’s legal responsibility for search responses.
🤖 Robotics
Woman arrested in Philadelphia for allegedly blocking a delivery robot
October 18, 2026
A woman was arrested in Philadelphia after repeatedly blocking the path of an Uber delivery robot and refusing to comply with police instructions. The incident highlights how autonomous delivery services bring not only technical issues but also questions of labor policy and social acceptance into public spaces.