Follow us on Google News
Major global companies, including Goldman Sachs, Nvidia and Palantir, are tightening employee access to artificial intelligence tools amid growing concerns over data security, confidential business information and rising technology costs.
The restrictions reflect a shift in corporate AI adoption, with organisations seeking greater control over how employees use advanced AI models in their daily work.
Goldman Sachs has restricted access to certain AI coding tools, while other companies are introducing controlled environments, stricter data protection requirements and spending limits.
Why Has Goldman Sachs Restricted AI Access?
Goldman Sachs has reportedly limited access to Anthropic’s Claude Code and Microsoft’s GitHub coding tools to engineers and other authorised employees. Previously, employees across departments, including bankers and traders, were encouraged to experiment with these technologies.
Traders had used AI tools to monitor client trading activity, analyse market movements and positions, and create written and graphic content.
https://the420.in/fcrf-certified-chief-ai-officer-second-cohort-november-2026/
Following an internal review of productivity and risk management, the bank decided to place the tools in the hands of employees considered better suited to their long-term use.
The restrictions do not apply to certain specialist employees working across trading, quantitative research and software engineering.
What Security Risks Are Companies Concerned About?
The primary concern involves protecting confidential information, proprietary software and sensitive internal systems.
Companies are reassessing whether external AI models could expose valuable business information or intellectual property.
Palantir has reportedly sought irrevocable zero-data-retention guarantees from Anthropic before making its models available through Palantir’s software.
Nvidia has restricted the use of Anthropic’s models to less-sensitive activities, including open-source software projects.
For sensitive internal operations, such as AI-driven supply chain monitoring, Nvidia uses its own Nemotron models.
Why Is Booz Allen Hamilton Restricting AI?
Defence contractor Booz Allen Hamilton has barred employees from using Anthropic’s AI models for work involving proprietary cybersecurity software developed for customers. The company previously used the technology to examine systems for potential cybersecurity vulnerabilities.
Its chief technology officer, Bill Vass, expressed concerns that external AI models might learn from the company’s proprietary code.
The restrictions reflect wider concerns about how commercially sensitive software and technical information are handled when employees use external AI services.
How Is JPMorgan Managing AI Security?
JPMorgan Chase is developing a controlled coding environment called Devspace for employees using Claude Code.
The environment is designed to separate AI tools from employee credentials, sensitive files and internal banking systems. It would also allow the bank to monitor activities involving the technology.
JPMorgan has additionally emphasised choosing appropriate AI models for different tasks rather than using expensive advanced models for routine assignments.
Goldman Sachs has developed an internal AI assistant combining models from OpenAI, Anthropic and Google.
Are Rising AI Costs Also Driving Restrictions?
Beyond cybersecurity, companies are introducing spending controls as AI usage becomes more expensive.
Uber Technologies has reportedly capped monthly spending on AI coding tools such as Claude Code and Cursor at $1,500 per employee per tool.
Walmart has moved from unlimited access to fixed token allocations for employees using its internally developed AI tool, Code Puppy.
Consulting firm Accenture has introduced a monthly allocation of $29 in centrally funded Claude Enterprise credits per employee, with additional usage subject to separate budgets.
What Does This Mean for Workplace AI?
The developments indicate that companies are moving away from unrestricted experimentation towards controlled AI deployment.
Businesses increasingly want employees to benefit from AI-powered productivity without exposing confidential information or creating uncontrolled expenses.
Financial institutions and technology companies are adopting different approaches, ranging from restricted access and internal AI environments to model-specific security policies.
The changes show that corporate AI adoption is increasingly being shaped by security requirements, cost management and oversight rather than access to technology alone.
