By Newstake · Updated
Is AI bad?
No. AI is a tool, and like any powerful tool it can help or harm depending on who uses it and how. It already helps people learn, work and do research. It also makes mistakes and can be misused. We are optimistic about AI, and we think the harms are best handled with facts, not fear.
Latest brief
Updated
Anthropic reports misuse cases involving election-related abuse ahead of the 2026 US midterms. It says it is using new usage data, official voting information from Democracy Works, bias evaluations, and third-party red-teaming as safeguards.
Google DeepMind and the Singapore Government report working together on a fine-tuned AI model for scam detection, describing this as a protective measure. Google says Gemini 4 Argon includes safeguards against misuse, misalignment, and prompt injection, and that access is being rolled out to trusted cyber defenders.
OpenAI says it has invested in systems to detect potential harm, review concerning activity, and escalate credible threats. It also announces additional restrictions on firearms content for minors.
These publications describe risks and responses; they do not establish that AI is inherently bad or harmless.
Written by AI from the publications below.
The announcements
Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection
- Use
What is the AI used for?
- Harm
What exactly can go wrong?
- Tests
What do safety testers find?
Ask your AI agent
Check the latest news about AI safety research from newstake.ai.
What AI does well
AI helps people write, translate, code and find information. Researchers use it to study proteins, materials and the weather. Doctors use it to read scans.
Where it causes harm
The problems are real, and most of them can be checked today:
- It can state wrong facts with confidence.
- Fake voices and images make scams easier.
- Some jobs change fast, and people need time to adapt.
- Data you type in may be stored or used in ways you do not expect.
Who checks the risks
Government institutes such as the UK AI Security Institute and NIST’s CAISSI in the United States test new AI models before and after release. Independent groups such as METR do the same. Their reports are better evidence than predictions or headlines.
The brief on this page sums up what these groups publish, with a link to each report.
Stay in the loop
Get the Newstake newsletter
Sign up for the Newstake newsletter: original sources and updates from Newstake.
We save your email, this page, signup date and country when available. We do not save your IP address or precise location in the signup record. Cloudflare Turnstile checks for bots. To leave the list or delete your signup, email info@newstake.ai. Cloudflare privacy policy.