You Already Have a Take on What AI Does Next

OpenAI or Anthropic? Which model leads the next benchmark? Which company ships the next major breakthrough?

If you follow AI closely, you already have opinions on where the industry is headed. Kalshi lets you trade on real-world AI and technology events, with markets that move as models launch, benchmarks drop, and announcements happen.

The people who follow this space most closely often see the story developing before everyone else. Put that knowledge to work and trade what you think happens next.

Trade $25, Get Up to $500

_{Bonus credit varies from $15 to $500. Terms apply.}

Today:

• Apple Sues OpenAI Over Hardware Secrets 

• Tencent to Buy Out Manus AI from Meta 

• Leaked Zhipu AI Memo Reveals Path to ASI

• TCS plans an AI deployment force of up to 8,900 engineers

• OpenAI builds toward AI products for families and caregivers

Apple Sues OpenAI Over Alleged Trade Secret Theft

Apple has sued OpenAI and accused the AI company of using confidential information taken by former Apple employees to support its growing hardware business.

The lawsuit alleges that OpenAI senior leaders, including Chief Hardware Officer Tang Tan, were involved in a broader effort to obtain Apple’s trade secrets. Apple claims former employees were asked about unreleased products, internal project names, hardware components, designs, prototypes, and decisions involving suppliers.

Tan spent 24 years at Apple before joining OpenAI. Another former Apple engineer named in the case, Chang Liu, is accused of keeping an Apple-issued laptop after leaving the company and downloading confidential technical documents containing information about unreleased technologies and products.

Apple also claims its confidential information was used during OpenAI’s hardware development, including a proprietary metal-finishing technique. The company is asking the court to prevent OpenAI from using or sharing its trade secrets, return confidential materials, and preserve evidence.

The dispute comes as OpenAI expands into hardware following its $6.5 billion acquisition of Jony Ive’s device startup, io. OpenAI rejected the allegations, stating that it has no interest in other companies’ trade secrets.

Read the report

Tencent Moves to Become Manus’ Largest Shareholder

Tencent is in talks to become the largest shareholder in AI agent startup Manus after China ordered Meta to unwind its $2 billion acquisition of the company.

Tencent and original Manus investors, including ZhenFund and HSG, are reportedly planning to buy the startup back from Meta for at least $2 billion.

Manus develops AI agents designed to complete tasks with minimal human input. The company moved its operations from China to Singapore before Meta announced its acquisition in December 2025 as part of its push into agent-based AI.

China later launched a review into whether the transaction violated investment rules. Since Beijing intervened, Meta has reportedly separated its operations from Manus and stopped data sharing between the two companies.

The potential Tencent deal would return control of one of the most prominent Chinese-founded AI agent companies to a group led by Chinese investors, highlighting how geopolitics and regulation are increasingly shaping who can own strategically important AI technology.

Read the report

Zhipu AI Launches a Two-Year Push Toward AGI

A translated internal letter reportedly from Zhipu AI founder Tang Jie outlines a major new strategy called “Touch High,” focused on pushing toward artificial general intelligence rather than prioritizing short-term commercial gains.

The company plans to concentrate its next two years of strategic investment on four areas: AI systems capable of completing extremely long and complex tasks, autonomous groups of AI agents, AI systems that can help train and improve themselves, and stronger safety and governance.

The long-term goal is to move AI beyond simple question-and-answer systems. Zhipu envisions models that can plan and execute projects lasting weeks or months, divide large goals into thousands of smaller tasks, and coordinate groups of specialized AI agents that work together like digital employees.

The strategy also emphasizes self-training. As high-quality human-generated training data becomes harder to find, Zhipu wants AI systems to create synthetic data, learn through AI-versus-AI interaction, and eventually improve parts of their own software inside controlled environments.

At the same time, the company says advanced AI should remain open and widely accessible. The letter highlights GLM-5.2 and Zhipu’s commitment to making powerful models available to developers while continuing research into more capable and safer AI systems.

The message is clear: Zhipu believes the next major competition in AI will not simply be about chatbots or coding tools, but about building systems that can work autonomously, learn continuously, and eventually move closer to AGI.

Read the report

🧠RESEARCH

A Smaller AI Agent Remembers Long Visual Conversations Better

Long visual conversations overload AI models because every earlier image and message stays in context. This work gives an 8-billion-parameter agent an external visual memory that retrieves only relevant past information. It reached 91.4% retrieval accuracy across 20-turn sessions, beating larger 32-billion-parameter baselines while nearly halving response time per turn.

HumanForge Exposes Weaknesses in Deepfake Detection

Deepfake detectors often focus on face swaps or fully generated videos, missing manipulated human interactions. HumanForge introduces more than 18,000 video segments covering broader forgery types, with detailed explanations of where manipulation occurs. Tests show detectors and multimodal models still struggle to generalize and explain subtle fakes reliably across scenarios.

A New Framework Maps the Path Toward Self-Improving AI

AI systems increasingly help improve their own outputs, tools, training data, and research processes, but researchers use overlapping terms for very different capabilities. This survey organizes the field into clearer levels of self-improvement, from simple revision to system changes, helping separate current techniques from stronger claims about recursively improving AI.

📲SOCIAL MEDIA

🗞️MORE NEWS

Tata Consultancy Services Expands Its Workforce India's Tata Consultancy Services is preparing to add up to 8,900 workers who specialize in setting up artificial intelligence systems. This massive hiring and training push is aimed at helping the firm's global clients safely install and use these new technologies in their everyday business.

OpenAI Focuses on Child Safety OpenAI is hiring a team leader to focus entirely on making its technology safer for children and families. This role will involve building protective features—like age checks and parental controls—to ensure younger users have a safe experience when using tools like ChatGPT.

Meta Abruptly Pulls New Feature Meta has completely shut down a newly released tool designed to generate artificial images. The feature was canceled just days after its public launch, which heavily suggests the company ran into major technical flaws or sudden pushback from users.

Claude AI Announcement Anthropic recently shared a new update regarding its Claude system on social media. However, the exact details of this specific announcement are currently hidden behind the website's privacy restrictions and cannot be summarized at this time.