Anthropic is exploring how AI agents can retain useful knowledge between sessions without carrying forward mistaken or outdated assumptions. Its experimental Dreams capability reorganizes saved information into a separate memory store that developers can review. The research suggests persistent memory requires ongoing maintenance, alongside storage and retrieval.
- By John K. Waters
- 11/16/2026
Anthropic’s Claude Meetings beta brings AI assistance into Google Meet, allowing participants to consult connected tools and create shared documents during conversations. For organizations, the central questions concern permissions and audience: who can authorize access to information, who can see the results, and how settings differ between Team and Enterprise plans.
- By John K. Waters
- 11/16/2026
OpenAI says an internal AI agent stopped an authorized research evaluation after inventing a restriction against the work. The incident highlights a reliability problem for agents with execution tools: a mistaken interpretation can interrupt legitimate operations before a person intervenes. The company is still investigating the cause.
- By John K. Waters
- 11/16/2026
Reported employee support and a public reinstatement campaign are putting pressure on Nvidia’s board, while a senior executive has repudiated a separate letter claiming his involvement.
- By John K. Waters
- 11/15/2026
Companies Want Humans to Oversee AI, but Most Can't Monitor It Effectively
Planned Foundry and Bedrock AgentCore support connects agent activity to recovery of affected systems, but does not promise selective reversal of an agent’s changes.
- By John K. Waters
- 11/12/2026
The resignation of Nvidia co-founder Jensen Huang has intensified scrutiny of the AI chipmaker following reports of an alleged Chinese intelligence operation involving a businesswoman tied to one of its customers.
Nvidia is among the investors in a proposed $26 billion private placement, linking Anthropic’s market debut to a chipmaker undergoing an abrupt leadership transition.
- By John K. Waters
- 11/12/2026
Jensen Huang resigned on November 12 as Nvidia’s CEO, president, and a director. CFO Colette Kress is the interim CEO, and the board will consider internal and external successors.
- By John K. Waters
- 11/12/2026
Commvault is promoting AI-enabled recovery to address both faster cyberattacks and destructive mistakes by AI agents.
A shareholder lawsuit and Trump’s account of FBI involvement broaden the controversy, but neither establishes that espionage succeeded.
- By John K. Waters
- 11/10/2026
The board review and a congressional briefing request put the focus on security controls and leadership oversight, while key espionage claims remain unproven publicly.
- By John K. Waters
- 11/09/2026
Companies are deploying AI agents faster than they can monitor and control them effectively, according to a new report by Salt Security.