|
KEP AI BULLETIN
Sunday 07 June 2026 |
Edition 006
8 stories across 5 sections today.
|
|
Models & Research
|
|
Govt agencies, private companies among Indian entities given access to Anthropic's Mythos
Anthropic has given select Indian government agencies and private companies access to its Mythos AI model as part of a program called Project Glasswing. The arrangement is focused on cybersecurity applications. Details on which specific organizations are involved or the scope of access have not been fully disclosed.
|
|
New Tools & Products
|
|
OpenAI Rolls Out Lockdown Mode to Fight Prompt Injection Attacks
OpenAI has introduced a 'Lockdown Mode' for its AI tools designed to block prompt injection attacks, where malicious instructions are hidden in content to manipulate the AI. The protection comes with a significant tradeoff: users lose access to features like live web browsing and retrieving images from the internet. Organizations handling sensitive data may find the security benefit worth the redu
|
|
Chips & Infrastructure
|
|
Bank of England governor warns AI may need to be rationed because of energy limits
Bank of England Governor Andrew Bailey has cautioned that AI adoption could outpace available energy supply, potentially forcing governments and companies to limit or ration its use. He described this as a major societal decision that policymakers will need to confront. The warning highlights growing concern that AI infrastructure demands may clash with energy capacity constraints.
|
|
Robotics & Physical AI
|
|
Chinese humanoid robots dominate the market with thousands shipped a year. But most are still performative rather than functional
Chinese companies are shipping thousands of humanoid robots annually, giving them a dominant position in the market. However, most of these robots are still being used for demonstrations rather than performing practical, real-world work tasks. Experts note that without stronger commercial demand, manufacturers cannot achieve the scale needed for true mass production.
|
|
AGIBOT holds World Challenge 2026 to see how AI models perform on real tasks
Robotics company AGIBOT is hosting a competition in 2026 that tests AI models on real physical tasks using actual robots, rather than relying on simulated environments. The event reflects a broader industry shift toward evaluating AI performance in real-world conditions. The goal is to produce more meaningful benchmarks for how AI systems handle practical work.
|
|
Business & Investment
|
|
Bernie Sanders and Sam Altman’s private one-hour meeting about the public ownership of AI
Senator Bernie Sanders held a private, hour-long meeting with OpenAI CEO Sam Altman to discuss the idea of public ownership of AI systems. Separately, President Trump indicated that major AI company executives are expected to visit the White House soon to talk about a potential government partnership. The conversations signal growing political interest in how AI ownership and governance should be
|
|
Donald Trump, Bernie Sanders and Sam Altman are all talking about public ownership in AI - AP News
Public ownership of AI has become a talking point across the political spectrum, with Donald Trump, Bernie Sanders, and Sam Altman all engaging with the idea in different ways. Sanders has proposed legislation that would require the government to take a 50% equity stake in companies like OpenAI and Anthropic. Trump is separately reported to be considering some form of government stake in leading A
|
|
Trump signs AI memo addressing issues in Anthropic-Pentagon feud
President Trump has signed a memo requiring national security agencies to work with multiple AI providers rather than relying on a single company. The directive follows a reported dispute between Anthropic and the Pentagon, and instructs agencies to avoid AI vendors that interfere with military chain-of-command. The policy is intended to prevent over-dependence on any one AI supplier in sensitive
|
|
KEP Insight
|
|
Anthropic quietly giving Indian government agencies and select private companies access to a specialised AI model for cybersecurity is worth paying attention to, because it signals that advanced AI is being wired into the country's security infrastructure at a foundational level. For SMEs, the practical read is this: the threat landscape you operate in is being shaped by tools your larger counterparts and regulators already have access to, which raises the bar on what counts as adequate cyber hygiene for smaller businesses. If you are still treating cybersecurity as an annual checkbox, this is a good moment to revisit that assumption and make sure your basic defences, access controls, and vendor risk practices are genuinely current.
— Prabhu Iyer, KEP
|
|
Knowledge Equity Partners
This bulletin is curated by KEP AI systems and reviewed for relevance. It is for informational purposes only.
|